Verified 3w ago Mainly llama.cpp 200 approved runs
Submissions
2,318
200 approved public runs
Models
16
unique models
Avg output
24.0tok/s
Best 103 tok/s
Avg prefill
72.7tok/s
Best 281 tok/s
Avg total
43.4tok/s
Best 215 tok/s
Avg TTFT
10,220ms
Best 1,110 ms
Hardware
2
distinct setups
Matrix runs
36
speed test matrices
Matrix runs36+
Matrix runs 36
A matrix run measures prompt reading (prefill) and generation (decode) speed in one go, across several context lengths and numbers of simultaneous requests. Showing the newest 10.
CONCURRENCY TEST, NOT A SINGLE-USER RESULT. 3 requests sent at once (lmx --concurrency 3); 16 tok/s is the aggregate decode rate across all 3 streams, about 5.3 tok/s per stream, and the 4311 ms time to first token is under that load. KASA cuda-vm: 1 x NVIDIA RTX A6000 48 GB (PCIe passthrough, driver 615.71.10) in a VM with 16 vCPU pinned to 8 of the 16 cores (16 threads) of an AMD EPYC 7343 (another VM holds the other 8 cores) and 128 GiB. llama.cpp llama-server b11889 (CUDA image ghcr.io/ggml-org/llama.cpp), placement by llama.cpp's default --fit (1 GiB margin): the 88 GB model does not fit the 48 GB GPU, so llama.cpp's --fit kept the layers on the GPU and placed part of the MoE expert weights in system RAM on the CPU (loaded with -lm none, so the CPU part is in ordinary RAM, not memory-mapped); the exact split is not logged at the default verbosity; 47488 MiB of VRAM in use after load, 8 generation threads, 16 batch threads, flash attention on, 8 server slots of 4096 tokens each, context 32768. GGUF unsloth/Qwen3-235B-A22B-Instruct-2507-GGUF @ 437d6915c5c512869e4cf8b16b840bbe3fb172bc, file Qwen3-235B-A22B-Instruct-2507-UD-Q2_K_XL-00001-of-00002.gguf. Each request: LocalMaxxing's canonical reasoning-v1 prompt with a unique nonce, 256 output tokens. One discarded warm-up batch, then 3 measured batches: 16, 22.4, 15 tok/s; this is the median. Not eligible for the Verified badge, which requires batch 1.
KASA b65-vm: 1 x Intel Arc Pro B65 32 GB (PCIe passthrough, xe driver, Ubuntu 26.04 kernel 7.0) in a VM with 16 vCPU pinned to 8 cores (16 threads) of an AMD EPYC 7343 and 128 GiB. Engine: XeStrata xe0.1.40.2.2 (MistVVK/XeStrata, the Intel fork of Strata; LGPL-3.0) in the chriscorbell/intel-arc-pro-b70-qwen38-flash-next recipe image built from commit 6aad3506 (local image local/xestrata-qwen38:xe0.1.40.2.2, sha256:6f03dde70323). Weights: ISTA-DASLab/Qwen3.8-Flash-Next-GSQ-RCO-GGUF IQ3_S @ ed59f920; the GPU holds the dense layers and a cache of hot experts, the CPU computes the rest of the experts from system RAM at the same time. MTP speculative decoding (4 tokens), int8 KV cache. Settings: SPEC_MIN_P=0.6 VRAM_RESERVE_MIB=1536 MAX_CONTEXT=65536 CONVERSATION_CACHE_MIB=4096. Canonical reasoning-v1, lmx single-shot requests after one discarded warm-up. Measured requests: 52.5, 53.5, 52.9 tok/s; this is the median (tie to the earlier). Campaign ai/vm/b65-vm/2026-10-27-b65-vm-localmaxxing. Prompt/output samples were added from the lmx run file; engineTimingsRaw is XeStrata's own per-request log (decode tokens and tok/s, expert-cache hit rate), because lmx does not capture it for this engine.
KASA b65-vm: 1 x Intel Arc Pro B65 32 GB (PCIe passthrough, xe driver, Ubuntu 26.04 kernel 7.0) in a VM with 16 vCPU pinned to 8 cores (16 threads) of an AMD EPYC 7343 and 128 GiB. Engine: XeStrata xe0.1.40.2.2 (MistVVK/XeStrata, the Intel fork of Strata; LGPL-3.0) in the chriscorbell/intel-arc-pro-b70-qwen38-flash-next recipe image built from commit 6aad3506 (local image local/xestrata-qwen38:xe0.1.40.2.2, sha256:6f03dde70323). Weights: ISTA-DASLab/Qwen3.8-Flash-Next-GSQ-RCO-GGUF IQ3_S @ ed59f920; the GPU holds the dense layers and a cache of hot experts, the CPU computes the rest of the experts from system RAM at the same time. MTP speculative decoding (4 tokens), int8 KV cache. Settings: SPEC_MIN_P=0.58 VRAM_RESERVE_MIB=1024 MAX_CONTEXT=65536 CONVERSATION_CACHE_MIB=2048. Canonical reasoning-v1, lmx single-shot requests after one discarded warm-up. Measured requests: 52.1, 57.9, 58.7 tok/s; this is the median (tie to the earlier). Campaign ai/vm/b65-vm/2026-10-27-b65-vm-localmaxxing. Prompt/output samples were added from the lmx run file; engineTimingsRaw is XeStrata's own per-request log (decode tokens and tok/s, expert-cache hit rate), because lmx does not capture it for this engine.
KASA b65-vm: 1 x Intel Arc Pro B65 32 GB (PCIe passthrough, xe driver, Ubuntu 26.04 kernel 7.0) in a VM with 16 vCPU pinned to 8 cores (16 threads) of an AMD EPYC 7343 and 128 GiB. Engine: XeStrata xe0.1.40.2.2 (MistVVK/XeStrata, the Intel fork of Strata; LGPL-3.0) in the chriscorbell/intel-arc-pro-b70-qwen38-flash-next recipe image built from commit 6aad3506 (local image local/xestrata-qwen38:xe0.1.40.2.2, sha256:6f03dde70323). Weights: ISTA-DASLab/Qwen3.8-Flash-Next-GSQ-RCO-GGUF IQ3_S @ ed59f920; the GPU holds the dense layers and a cache of hot experts, the CPU computes the rest of the experts from system RAM at the same time. MTP speculative decoding (4 tokens), int8 KV cache. Settings: SPEC_MIN_P=0.58 VRAM_RESERVE_MIB=1536 MAX_CONTEXT=65536 CONVERSATION_CACHE_MIB=4096. Canonical reasoning-v1, lmx single-shot requests after one discarded warm-up. Measured requests: 54.2, 55.3, 55.7 tok/s; this is the median (tie to the earlier). Campaign ai/vm/b65-vm/2026-10-27-b65-vm-localmaxxing. Prompt/output samples were added from the lmx run file; engineTimingsRaw is XeStrata's own per-request log (decode tokens and tok/s, expert-cache hit rate), because lmx does not capture it for this engine.
CONCURRENCY TEST, NOT A SINGLE-USER RESULT. 2 requests sent at once (lmx --concurrency 2); 20.5 tok/s is the aggregate decode rate across all 2 streams, about 10.2 tok/s per stream, and the 2329 ms time to first token is under that load. KASA cuda-vm: 1 x NVIDIA RTX A6000 48 GB (PCIe passthrough, driver 615.71.10) in a VM with 16 vCPU pinned to 8 of the 16 cores (16 threads) of an AMD EPYC 7343 (another VM holds the other 8 cores) and 128 GiB. llama.cpp llama-server b11889 (CUDA image ghcr.io/ggml-org/llama.cpp), placement by llama.cpp's default --fit (1 GiB margin): the 88 GB model does not fit the 48 GB GPU, so llama.cpp's --fit kept the layers on the GPU and placed part of the MoE expert weights in system RAM on the CPU (loaded with -lm none, so the CPU part is in ordinary RAM, not memory-mapped); the exact split is not logged at the default verbosity; 47488 MiB of VRAM in use after load, 8 generation threads, 16 batch threads, flash attention on, 8 server slots of 4096 tokens each, context 32768. GGUF unsloth/Qwen3-235B-A22B-Instruct-2507-GGUF @ 437d6915c5c512869e4cf8b16b840bbe3fb172bc, file Qwen3-235B-A22B-Instruct-2507-UD-Q2_K_XL-00001-of-00002.gguf. Each request: LocalMaxxing's canonical reasoning-v1 prompt with a unique nonce, 256 output tokens. One discarded warm-up batch, then 3 measured batches: 18.7, 20.5, 22 tok/s; this is the median. Not eligible for the Verified badge, which requires batch 1. Selection rule applied to every run in this round: the median of the 2 measured requests whose lmx promptTokens is a whole number; the others report a fractional median, which LocalMaxxing refuses.
KASA b65-vm: 1 x Intel Arc Pro B65 32 GB (PCIe passthrough, xe driver, Ubuntu 26.04 kernel 7.0) in a VM with 16 vCPU pinned to 8 cores (16 threads) of an AMD EPYC 7343 and 128 GiB. Engine: XeStrata xe0.1.40.2.2 (MistVVK/XeStrata, the Intel fork of Strata; LGPL-3.0) in the chriscorbell/intel-arc-pro-b70-qwen38-flash-next recipe image built from commit 6aad3506 (local image local/xestrata-qwen38:xe0.1.40.2.2, sha256:6f03dde70323). Weights: ISTA-DASLab/Qwen3.8-Flash-Next-GSQ-RCO-GGUF IQ3_S @ ed59f920; the GPU holds the dense layers and a cache of hot experts, the CPU computes the rest of the experts from system RAM at the same time. MTP speculative decoding (4 tokens), int8 KV cache. Settings: SPEC_MIN_P=0.62 VRAM_RESERVE_MIB=1280 MAX_CONTEXT=65536 CONVERSATION_CACHE_MIB=2048. Canonical reasoning-v1, lmx single-shot requests after one discarded warm-up. Measured requests: 54.4, 61.1, 58.2 tok/s; this is the median (tie to the earlier). Campaign ai/vm/b65-vm/2026-10-27-b65-vm-localmaxxing. Prompt/output samples were added from the lmx run file; engineTimingsRaw is XeStrata's own per-request log (decode tokens and tok/s, expert-cache hit rate), because lmx does not capture it for this engine.
KASA cuda-vm: 1 x NVIDIA RTX A6000 48 GB (PCIe passthrough, driver 615.71.10) in a VM with 16 vCPU pinned to 8 of the 16 cores (16 threads) of an AMD EPYC 7343 (another VM holds the other 8 cores) and 128 GiB. llama.cpp llama-server b11889 (CUDA image ghcr.io/ggml-org/llama.cpp), placement by llama.cpp's default --fit (1 GiB margin): the 88 GB model does not fit the 48 GB GPU, so llama.cpp's --fit kept the layers on the GPU and placed part of the MoE expert weights in system RAM on the CPU (loaded with -lm none, so the CPU part is in ordinary RAM, not memory-mapped); the exact split is not logged at the default verbosity; 47488 MiB of VRAM in use after load, 8 generation threads, 16 batch threads, flash attention on, 8 server slots of 4096 tokens each, context 32768. GGUF unsloth/Qwen3-235B-A22B-Instruct-2507-GGUF @ 437d6915c5c512869e4cf8b16b840bbe3fb172bc, file Qwen3-235B-A22B-Instruct-2507-UD-Q2_K_XL-00001-of-00002.gguf. Canonical reasoning-v1, lmx single-shot requests after one discarded warm-up. Measured requests: 14.1, 14.9, 12.8 tok/s; this is the median (tie to the earlier). Prompt and output samples were added from the lmx run file, engineTimingsRaw from exact llama-server /metrics deltas around the request.
KASA b65-vm: 1 x Intel Arc Pro B65 32 GB (PCIe passthrough, xe driver, Ubuntu 26.04 kernel 7.0) in a VM with 16 vCPU pinned to 8 cores (16 threads) of an AMD EPYC 7343 and 128 GiB. Engine: XeStrata xe0.1.40.2.2 (MistVVK/XeStrata, the Intel fork of Strata; LGPL-3.0) in the chriscorbell/intel-arc-pro-b70-qwen38-flash-next recipe image built from commit 6aad3506 (local image local/xestrata-qwen38:xe0.1.40.2.2, sha256:6f03dde70323). Weights: ISTA-DASLab/Qwen3.8-Flash-Next-GSQ-RCO-GGUF IQ3_S @ ed59f920; the GPU holds the dense layers and a cache of hot experts, the CPU computes the rest of the experts from system RAM at the same time. MTP speculative decoding (4 tokens), int8 KV cache. Settings: SPEC_MIN_P=0.58 VRAM_RESERVE_MIB=1024 MAX_CONTEXT=131072 CONVERSATION_CACHE_MIB=4096. Canonical reasoning-v1, lmx single-shot requests after one discarded warm-up. Measured requests: 54, 57.2, 59.5 tok/s; this is the median (tie to the earlier). Campaign ai/vm/b65-vm/2026-10-27-b65-vm-localmaxxing. Prompt/output samples were added from the lmx run file; engineTimingsRaw is XeStrata's own per-request log (decode tokens and tok/s, expert-cache hit rate), because lmx does not capture it for this engine.
KASA b65-vm: 1 x Intel Arc Pro B65 32 GB (PCIe passthrough, xe driver, Ubuntu 26.04 kernel 7.0) in a VM with 16 vCPU pinned to 8 cores (16 threads) of an AMD EPYC 7343 and 128 GiB. Engine: XeStrata xe0.1.40.2.2 (MistVVK/XeStrata, the Intel fork of Strata; LGPL-3.0) in the chriscorbell/intel-arc-pro-b70-qwen38-flash-next recipe image built from commit 6aad3506 (local image local/xestrata-qwen38:xe0.1.40.2.2, sha256:6f03dde70323). Weights: ISTA-DASLab/Qwen3.8-Flash-Next-GSQ-RCO-GGUF IQ3_S @ ed59f920; the GPU holds the dense layers and a cache of hot experts, the CPU computes the rest of the experts from system RAM at the same time. MTP speculative decoding (4 tokens), int8 KV cache. Settings: SPEC_MIN_P=0.58 VRAM_RESERVE_MIB=1024 MAX_CONTEXT=131072 CONVERSATION_CACHE_MIB=2048. Canonical reasoning-v1, lmx single-shot requests after one discarded warm-up. Measured requests: 40, 54, 55.8 tok/s; this is the median (tie to the earlier). Campaign ai/vm/b65-vm/2026-10-27-b65-vm-localmaxxing. Prompt/output samples were added from the lmx run file; engineTimingsRaw is XeStrata's own per-request log (decode tokens and tok/s, expert-cache hit rate), because lmx does not capture it for this engine.
KASA b65-vm: 1 x Intel Arc Pro B65 32 GB (PCIe passthrough, xe driver, Ubuntu 26.04 kernel 7.0) in a VM with 16 vCPU pinned to 8 cores (16 threads) of an AMD EPYC 7343 and 128 GiB. Engine: XeStrata xe0.1.40.2.2 (MistVVK/XeStrata, the Intel fork of Strata; LGPL-3.0) in the chriscorbell/intel-arc-pro-b70-qwen38-flash-next recipe image built from commit 6aad3506 (local image local/xestrata-qwen38:xe0.1.40.2.2, sha256:6f03dde70323). Weights: ISTA-DASLab/Qwen3.8-Flash-Next-GSQ-RCO-GGUF IQ3_S @ ed59f920; the GPU holds the dense layers and a cache of hot experts, the CPU computes the rest of the experts from system RAM at the same time. MTP speculative decoding (4 tokens), int8 KV cache. Settings: SPEC_MIN_P=0.58 VRAM_RESERVE_MIB=1536 MAX_CONTEXT=65536 CONVERSATION_CACHE_MIB=2048. Canonical reasoning-v1, lmx single-shot requests after one discarded warm-up. Measured requests: 51, 57.6, 60.3 tok/s; this is the median (tie to the earlier). Campaign ai/vm/b65-vm/2026-10-27-b65-vm-localmaxxing. Prompt/output samples were added from the lmx run file; engineTimingsRaw is XeStrata's own per-request log (decode tokens and tok/s, expert-cache hit rate), because lmx does not capture it for this engine.
KASA b65-vm: 1 x Intel Arc Pro B65 32 GB (PCIe passthrough, xe driver, Ubuntu 26.04 kernel 7.0) in a VM with 16 vCPU pinned to 8 cores (16 threads) of an AMD EPYC 7343 and 128 GiB. Engine: XeStrata xe0.1.40.2.2 (MistVVK/XeStrata, the Intel fork of Strata; LGPL-3.0) in the chriscorbell/intel-arc-pro-b70-qwen38-flash-next recipe image built from commit 6aad3506 (local image local/xestrata-qwen38:xe0.1.40.2.2, sha256:6f03dde70323). Weights: ISTA-DASLab/Qwen3.8-Flash-Next-GSQ-RCO-GGUF IQ3_S @ ed59f920; the GPU holds the dense layers and a cache of hot experts, the CPU computes the rest of the experts from system RAM at the same time. MTP speculative decoding (4 tokens), int8 KV cache. Settings: SPEC_MIN_P=0.6 VRAM_RESERVE_MIB=1536 MAX_CONTEXT=131072 CONVERSATION_CACHE_MIB=4096. Canonical reasoning-v1, lmx single-shot requests after one discarded warm-up. Measured requests: 51.8, 54, 55.4 tok/s; this is the median (tie to the earlier). Campaign ai/vm/b65-vm/2026-10-27-b65-vm-localmaxxing. Prompt/output samples were added from the lmx run file; engineTimingsRaw is XeStrata's own per-request log (decode tokens and tok/s, expert-cache hit rate), because lmx does not capture it for this engine.
KASA b65-vm: 1 x Intel Arc Pro B65 32 GB (PCIe passthrough, xe driver, Ubuntu 26.04 kernel 7.0) in a VM with 16 vCPU pinned to 8 cores (16 threads) of an AMD EPYC 7343 and 128 GiB. Engine: XeStrata xe0.1.40.2.2 (MistVVK/XeStrata, the Intel fork of Strata; LGPL-3.0) in the chriscorbell/intel-arc-pro-b70-qwen38-flash-next recipe image built from commit 6aad3506 (local image local/xestrata-qwen38:xe0.1.40.2.2, sha256:6f03dde70323). Weights: ISTA-DASLab/Qwen3.8-Flash-Next-GSQ-RCO-GGUF IQ3_S @ ed59f920; the GPU holds the dense layers and a cache of hot experts, the CPU computes the rest of the experts from system RAM at the same time. MTP speculative decoding (4 tokens), int8 KV cache. Settings: SPEC_MIN_P=0.58 VRAM_RESERVE_MIB=1536 MAX_CONTEXT=131072 CONVERSATION_CACHE_MIB=2048. Canonical reasoning-v1, lmx single-shot requests after one discarded warm-up. Measured requests: 50.7, 56.8, 54.8 tok/s; this is the median (tie to the earlier). Campaign ai/vm/b65-vm/2026-10-27-b65-vm-localmaxxing. Prompt/output samples were added from the lmx run file; engineTimingsRaw is XeStrata's own per-request log (decode tokens and tok/s, expert-cache hit rate), because lmx does not capture it for this engine.
KASA b65-vm: 1 x Intel Arc Pro B65 32 GB (PCIe passthrough, xe driver, Ubuntu 26.04 kernel 7.0) in a VM with 16 vCPU pinned to 8 cores (16 threads) of an AMD EPYC 7343 and 128 GiB. Engine: XeStrata xe0.1.40.2.2 (MistVVK/XeStrata, the Intel fork of Strata; LGPL-3.0) in the chriscorbell/intel-arc-pro-b70-qwen38-flash-next recipe image built from commit 6aad3506 (local image local/xestrata-qwen38:xe0.1.40.2.2, sha256:6f03dde70323). Weights: ISTA-DASLab/Qwen3.8-Flash-Next-GSQ-RCO-GGUF IQ3_S @ ed59f920; the GPU holds the dense layers and a cache of hot experts, the CPU computes the rest of the experts from system RAM at the same time. MTP speculative decoding (4 tokens), int8 KV cache. Settings: SPEC_MIN_P=0.58 VRAM_RESERVE_MIB=1280 MAX_CONTEXT=131072 CONVERSATION_CACHE_MIB=4096. Canonical reasoning-v1, lmx single-shot requests after one discarded warm-up. Measured requests: 50.5, 57.6, 59.4 tok/s; this is the median (tie to the earlier). Campaign ai/vm/b65-vm/2026-10-27-b65-vm-localmaxxing. Prompt/output samples were added from the lmx run file; engineTimingsRaw is XeStrata's own per-request log (decode tokens and tok/s, expert-cache hit rate), because lmx does not capture it for this engine.
KASA b65-vm: 1 x Intel Arc Pro B65 32 GB (PCIe passthrough, xe driver, Ubuntu 26.04 kernel 7.0) in a VM with 16 vCPU pinned to 8 cores (16 threads) of an AMD EPYC 7343 and 128 GiB. Engine: XeStrata xe0.1.40.2.2 (MistVVK/XeStrata, the Intel fork of Strata; LGPL-3.0) in the chriscorbell/intel-arc-pro-b70-qwen38-flash-next recipe image built from commit 6aad3506 (local image local/xestrata-qwen38:xe0.1.40.2.2, sha256:6f03dde70323). Weights: ISTA-DASLab/Qwen3.8-Flash-Next-GSQ-RCO-GGUF IQ3_S @ ed59f920; the GPU holds the dense layers and a cache of hot experts, the CPU computes the rest of the experts from system RAM at the same time. MTP speculative decoding (4 tokens), int8 KV cache. Settings: SPEC_MIN_P=0.62 VRAM_RESERVE_MIB=1536 MAX_CONTEXT=131072 CONVERSATION_CACHE_MIB=2048. Canonical reasoning-v1, lmx single-shot requests after one discarded warm-up. Measured requests: 54.8, 56, 56.3 tok/s; this is the median (tie to the earlier). Campaign ai/vm/b65-vm/2026-10-27-b65-vm-localmaxxing. Prompt/output samples were added from the lmx run file; engineTimingsRaw is XeStrata's own per-request log (decode tokens and tok/s, expert-cache hit rate), because lmx does not capture it for this engine.
KASA b65-vm: 1 x Intel Arc Pro B65 32 GB (PCIe passthrough, xe driver, Ubuntu 26.04 kernel 7.0) in a VM with 16 vCPU pinned to 8 cores (16 threads) of an AMD EPYC 7343 and 128 GiB. Engine: XeStrata xe0.1.40.2.2 (MistVVK/XeStrata, the Intel fork of Strata; LGPL-3.0) in the chriscorbell/intel-arc-pro-b70-qwen38-flash-next recipe image built from commit 6aad3506 (local image local/xestrata-qwen38:xe0.1.40.2.2, sha256:6f03dde70323). Weights: ISTA-DASLab/Qwen3.8-Flash-Next-GSQ-RCO-GGUF IQ3_S @ ed59f920; the GPU holds the dense layers and a cache of hot experts, the CPU computes the rest of the experts from system RAM at the same time. MTP speculative decoding (4 tokens), int8 KV cache. Settings: SPEC_MIN_P=0.6 VRAM_RESERVE_MIB=1280 MAX_CONTEXT=131072 CONVERSATION_CACHE_MIB=2048. Canonical reasoning-v1, lmx single-shot requests after one discarded warm-up. Measured requests: 53.3, 58, 55 tok/s; this is the median (tie to the earlier). Campaign ai/vm/b65-vm/2026-10-27-b65-vm-localmaxxing. Prompt/output samples were added from the lmx run file; engineTimingsRaw is XeStrata's own per-request log (decode tokens and tok/s, expert-cache hit rate), because lmx does not capture it for this engine.
KASA b65-vm: 1 x Intel Arc Pro B65 32 GB (PCIe passthrough, xe driver, Ubuntu 26.04 kernel 7.0) in a VM with 16 vCPU pinned to 8 cores (16 threads) of an AMD EPYC 7343 and 128 GiB. Engine: XeStrata xe0.1.40.2.2 (MistVVK/XeStrata, the Intel fork of Strata; LGPL-3.0) in the chriscorbell/intel-arc-pro-b70-qwen38-flash-next recipe image built from commit 6aad3506 (local image local/xestrata-qwen38:xe0.1.40.2.2, sha256:6f03dde70323). Weights: ISTA-DASLab/Qwen3.8-Flash-Next-GSQ-RCO-GGUF IQ3_S @ ed59f920; the GPU holds the dense layers and a cache of hot experts, the CPU computes the rest of the experts from system RAM at the same time. MTP speculative decoding (4 tokens), int8 KV cache. Settings: SPEC_MIN_P=0.62 VRAM_RESERVE_MIB=1280 MAX_CONTEXT=131072 CONVERSATION_CACHE_MIB=4096. Canonical reasoning-v1, lmx single-shot requests after one discarded warm-up. Measured requests: 51.7, 59.6, 56.3 tok/s; this is the median (tie to the earlier). Campaign ai/vm/b65-vm/2026-10-27-b65-vm-localmaxxing. Prompt/output samples were added from the lmx run file; engineTimingsRaw is XeStrata's own per-request log (decode tokens and tok/s, expert-cache hit rate), because lmx does not capture it for this engine.
KASA b65-vm: 1 x Intel Arc Pro B65 32 GB (PCIe passthrough, xe driver, Ubuntu 26.04 kernel 7.0) in a VM with 16 vCPU pinned to 8 cores (16 threads) of an AMD EPYC 7343 and 128 GiB. Engine: XeStrata xe0.1.40.2.2 (MistVVK/XeStrata, the Intel fork of Strata; LGPL-3.0) in the chriscorbell/intel-arc-pro-b70-qwen38-flash-next recipe image built from commit 6aad3506 (local image local/xestrata-qwen38:xe0.1.40.2.2, sha256:6f03dde70323). Weights: ISTA-DASLab/Qwen3.8-Flash-Next-GSQ-RCO-GGUF IQ3_S @ ed59f920; the GPU holds the dense layers and a cache of hot experts, the CPU computes the rest of the experts from system RAM at the same time. MTP speculative decoding (4 tokens), int8 KV cache. Settings: SPEC_MIN_P=0.62 VRAM_RESERVE_MIB=1536 MAX_CONTEXT=131072 CONVERSATION_CACHE_MIB=4096. Canonical reasoning-v1, lmx single-shot requests after one discarded warm-up. Measured requests: 54.7, 57.5, 58.1 tok/s; this is the median (tie to the earlier). Campaign ai/vm/b65-vm/2026-10-27-b65-vm-localmaxxing. Prompt/output samples were added from the lmx run file; engineTimingsRaw is XeStrata's own per-request log (decode tokens and tok/s, expert-cache hit rate), because lmx does not capture it for this engine.
KASA b65-vm: 1 x Intel Arc Pro B65 32 GB (PCIe passthrough, xe driver, Ubuntu 26.04 kernel 7.0) in a VM with 16 vCPU pinned to 8 cores (16 threads) of an AMD EPYC 7343 and 128 GiB. Engine: XeStrata xe0.1.40.2.2 (MistVVK/XeStrata, the Intel fork of Strata; LGPL-3.0) in the chriscorbell/intel-arc-pro-b70-qwen38-flash-next recipe image built from commit 6aad3506 (local image local/xestrata-qwen38:xe0.1.40.2.2, sha256:6f03dde70323). Weights: ISTA-DASLab/Qwen3.8-Flash-Next-GSQ-RCO-GGUF IQ3_S @ ed59f920; the GPU holds the dense layers and a cache of hot experts, the CPU computes the rest of the experts from system RAM at the same time. MTP speculative decoding (4 tokens), int8 KV cache. Settings: SPEC_MIN_P=0.6 VRAM_RESERVE_MIB=1280 MAX_CONTEXT=131072 CONVERSATION_CACHE_MIB=4096. Canonical reasoning-v1, lmx single-shot requests after one discarded warm-up. Measured requests: 51.8, 58.4, 56.5 tok/s; this is the median (tie to the earlier). Campaign ai/vm/b65-vm/2026-10-27-b65-vm-localmaxxing. Prompt/output samples were added from the lmx run file; engineTimingsRaw is XeStrata's own per-request log (decode tokens and tok/s, expert-cache hit rate), because lmx does not capture it for this engine.
CONCURRENCY TEST, NOT A SINGLE-USER RESULT. 6 requests sent at once (lmx --concurrency 6); 5.1 tok/s is the aggregate decode rate across all 6 streams, about 0.8 tok/s per stream, and the 12149 ms time to first token is under that load. KASA cuda-vm: 1 x NVIDIA RTX A6000 48 GB (PCIe passthrough, driver 615.71.10) in a VM with 16 vCPU pinned to 8 of the 16 cores (16 threads) of an AMD EPYC 7343 (another VM holds the other 8 cores) and 128 GiB. llama.cpp llama-server b11889 (CUDA image ghcr.io/ggml-org/llama.cpp), placement by llama.cpp's default --fit (1 GiB margin): the 122 GB model does not fit the 48 GB GPU, so llama.cpp's --fit kept the layers on the GPU and placed part of the MoE expert weights in system RAM on the CPU (loaded with -lm none, so the CPU part is in ordinary RAM, not memory-mapped); the exact split is not logged at the default verbosity; 47658 MiB of VRAM in use after load, 8 generation threads, 16 batch threads, flash attention on, 8 server slots of 4096 tokens each, context 32768. GGUF unsloth/GLM-4.7-GGUF @ 70deda3fcbe622671ff2111d7c0c7392f618e3b0, file GLM-4.7-UD-IQ2_M-00001-of-00003.gguf. Each request: LocalMaxxing's canonical reasoning-v1 prompt with a unique nonce, 256 output tokens. One discarded warm-up batch, then 3 measured batches: 3.4, 5.1, 3.8 tok/s; this is the median. Not eligible for the Verified badge, which requires batch 1. Selection rule applied to every run in this round: the median of the 1 measured requests whose lmx promptTokens is a whole number; the others report a fractional median, which LocalMaxxing refuses.
KASA b65-vm: 1 x Intel Arc Pro B65 32 GB (PCIe passthrough, xe driver, Ubuntu 26.04 kernel 7.0) in a VM with 16 vCPU pinned to 8 cores (16 threads) of an AMD EPYC 7343 and 128 GiB. Engine: XeStrata xe0.1.40.2.2 (MistVVK/XeStrata, the Intel fork of Strata; LGPL-3.0) in the chriscorbell/intel-arc-pro-b70-qwen38-flash-next recipe image built from commit 6aad3506 (local image local/xestrata-qwen38:xe0.1.40.2.2, sha256:6f03dde70323). Weights: ISTA-DASLab/Qwen3.8-Flash-Next-GSQ-RCO-GGUF IQ3_S @ ed59f920; the GPU holds the dense layers and a cache of hot experts, the CPU computes the rest of the experts from system RAM at the same time. MTP speculative decoding (4 tokens), int8 KV cache. Settings: SPEC_MIN_P=0.58 VRAM_RESERVE_MIB=1280 MAX_CONTEXT=65536 CONVERSATION_CACHE_MIB=4096. Canonical reasoning-v1, lmx single-shot requests after one discarded warm-up. Measured requests: 46.4, 51.2, 53.2 tok/s; this is the median (tie to the earlier). Campaign ai/vm/b65-vm/2026-10-27-b65-vm-localmaxxing. Prompt/output samples were added from the lmx run file; engineTimingsRaw is XeStrata's own per-request log (decode tokens and tok/s, expert-cache hit rate), because lmx does not capture it for this engine.
KASA b65-vm: 1 x Intel Arc Pro B65 32 GB (PCIe passthrough, xe driver, Ubuntu 26.04 kernel 7.0) in a VM with 16 vCPU pinned to 8 cores (16 threads) of an AMD EPYC 7343 and 128 GiB. Engine: XeStrata xe0.1.40.2.2 (MistVVK/XeStrata, the Intel fork of Strata; LGPL-3.0) in the chriscorbell/intel-arc-pro-b70-qwen38-flash-next recipe image built from commit 6aad3506 (local image local/xestrata-qwen38:xe0.1.40.2.2, sha256:6f03dde70323). Weights: ISTA-DASLab/Qwen3.8-Flash-Next-GSQ-RCO-GGUF IQ3_S @ ed59f920; the GPU holds the dense layers and a cache of hot experts, the CPU computes the rest of the experts from system RAM at the same time. MTP speculative decoding (4 tokens), int8 KV cache. Settings: SPEC_MIN_P=0.62 VRAM_RESERVE_MIB=1024 MAX_CONTEXT=131072 CONVERSATION_CACHE_MIB=2048. Canonical reasoning-v1, lmx single-shot requests after one discarded warm-up. Measured requests: 48.2, 54.4, 57.5 tok/s; this is the median (tie to the earlier). Campaign ai/vm/b65-vm/2026-10-27-b65-vm-localmaxxing. Prompt/output samples were added from the lmx run file; engineTimingsRaw is XeStrata's own per-request log (decode tokens and tok/s, expert-cache hit rate), because lmx does not capture it for this engine.
KASA b65-vm: 1 x Intel Arc Pro B65 32 GB (PCIe passthrough, xe driver, Ubuntu 26.04 kernel 7.0) in a VM with 16 vCPU pinned to 8 cores (16 threads) of an AMD EPYC 7343 and 128 GiB. Engine: XeStrata xe0.1.40.2.2 (MistVVK/XeStrata, the Intel fork of Strata; LGPL-3.0) in the chriscorbell/intel-arc-pro-b70-qwen38-flash-next recipe image built from commit 6aad3506 (local image local/xestrata-qwen38:xe0.1.40.2.2, sha256:6f03dde70323). Weights: ISTA-DASLab/Qwen3.8-Flash-Next-GSQ-RCO-GGUF IQ3_S @ ed59f920; the GPU holds the dense layers and a cache of hot experts, the CPU computes the rest of the experts from system RAM at the same time. MTP speculative decoding (4 tokens), int8 KV cache. Settings: SPEC_MIN_P=0.62 VRAM_RESERVE_MIB=1536 MAX_CONTEXT=65536 CONVERSATION_CACHE_MIB=4096. Canonical reasoning-v1, lmx single-shot requests after one discarded warm-up. Measured requests: 49.3, 56.6, 55.6 tok/s; this is the median (tie to the earlier). Campaign ai/vm/b65-vm/2026-10-27-b65-vm-localmaxxing. Prompt/output samples were added from the lmx run file; engineTimingsRaw is XeStrata's own per-request log (decode tokens and tok/s, expert-cache hit rate), because lmx does not capture it for this engine.
KASA b65-vm: 1 x Intel Arc Pro B65 32 GB (PCIe passthrough, xe driver, Ubuntu 26.04 kernel 7.0) in a VM with 16 vCPU pinned to 8 cores (16 threads) of an AMD EPYC 7343 and 128 GiB. Engine: XeStrata xe0.1.40.2.2 (MistVVK/XeStrata, the Intel fork of Strata; LGPL-3.0) in the chriscorbell/intel-arc-pro-b70-qwen38-flash-next recipe image built from commit 6aad3506 (local image local/xestrata-qwen38:xe0.1.40.2.2, sha256:6f03dde70323). Weights: ISTA-DASLab/Qwen3.8-Flash-Next-GSQ-RCO-GGUF IQ3_S @ ed59f920; the GPU holds the dense layers and a cache of hot experts, the CPU computes the rest of the experts from system RAM at the same time. MTP speculative decoding (4 tokens), int8 KV cache. Settings: SPEC_MIN_P=0.62 VRAM_RESERVE_MIB=1024 MAX_CONTEXT=131072 CONVERSATION_CACHE_MIB=4096. Canonical reasoning-v1, lmx single-shot requests after one discarded warm-up. Measured requests: 53.4, 57.5, 60.3 tok/s; this is the median (tie to the earlier). Campaign ai/vm/b65-vm/2026-10-27-b65-vm-localmaxxing. Prompt/output samples were added from the lmx run file; engineTimingsRaw is XeStrata's own per-request log (decode tokens and tok/s, expert-cache hit rate), because lmx does not capture it for this engine.
KASA b65-vm: 1 x Intel Arc Pro B65 32 GB (PCIe passthrough, xe driver, Ubuntu 26.04 kernel 7.0) in a VM with 16 vCPU pinned to 8 cores (16 threads) of an AMD EPYC 7343 and 128 GiB. Engine: XeStrata xe0.1.40.2.2 (MistVVK/XeStrata, the Intel fork of Strata; LGPL-3.0) in the chriscorbell/intel-arc-pro-b70-qwen38-flash-next recipe image built from commit 6aad3506 (local image local/xestrata-qwen38:xe0.1.40.2.2, sha256:6f03dde70323). Weights: ISTA-DASLab/Qwen3.8-Flash-Next-GSQ-RCO-GGUF IQ3_S @ ed59f920; the GPU holds the dense layers and a cache of hot experts, the CPU computes the rest of the experts from system RAM at the same time. MTP speculative decoding (4 tokens), int8 KV cache. Settings: SPEC_MIN_P=0.58 VRAM_RESERVE_MIB=1280 MAX_CONTEXT=65536 CONVERSATION_CACHE_MIB=2048. Canonical reasoning-v1, lmx single-shot requests after one discarded warm-up. Measured requests: 50.4, 55.4, 60.1 tok/s; this is the median (tie to the earlier). Campaign ai/vm/b65-vm/2026-10-27-b65-vm-localmaxxing. Prompt/output samples were added from the lmx run file; engineTimingsRaw is XeStrata's own per-request log (decode tokens and tok/s, expert-cache hit rate), because lmx does not capture it for this engine.
KASA b65-vm: 1 x Intel Arc Pro B65 32 GB (PCIe passthrough, xe driver, Ubuntu 26.04 kernel 7.0) in a VM with 16 vCPU pinned to 8 cores (16 threads) of an AMD EPYC 7343 and 128 GiB. Engine: XeStrata xe0.1.40.2.2 (MistVVK/XeStrata, the Intel fork of Strata; LGPL-3.0) in the chriscorbell/intel-arc-pro-b70-qwen38-flash-next recipe image built from commit 6aad3506 (local image local/xestrata-qwen38:xe0.1.40.2.2, sha256:6f03dde70323). Weights: ISTA-DASLab/Qwen3.8-Flash-Next-GSQ-RCO-GGUF IQ3_S @ ed59f920; the GPU holds the dense layers and a cache of hot experts, the CPU computes the rest of the experts from system RAM at the same time. MTP speculative decoding (4 tokens), int8 KV cache. Settings: SPEC_MIN_P=0.6 VRAM_RESERVE_MIB=1280 MAX_CONTEXT=65536 CONVERSATION_CACHE_MIB=4096. Canonical reasoning-v1, lmx single-shot requests after one discarded warm-up. Measured requests: 53, 57.8, 55.1 tok/s; this is the median (tie to the earlier). Campaign ai/vm/b65-vm/2026-10-27-b65-vm-localmaxxing. Prompt/output samples were added from the lmx run file; engineTimingsRaw is XeStrata's own per-request log (decode tokens and tok/s, expert-cache hit rate), because lmx does not capture it for this engine.
CONCURRENCY TEST, NOT A SINGLE-USER RESULT. 3 requests sent at once (lmx --concurrency 3); 16 tok/s is the aggregate decode rate across all 3 streams, about 5.3 tok/s per stream, and the 4311 ms time to first token is under that load. KASA cuda-vm: 1 x NVIDIA RTX A6000 48 GB (PCIe passthrough, driver 615.71.10) in a VM with 16 vCPU pinned to 8 of the 16 cores (16 threads) of an AMD EPYC 7343 (another VM holds the other 8 cores) and 128 GiB. llama.cpp llama-server b11889 (CUDA image ghcr.io/ggml-org/llama.cpp), placement by llama.cpp's default --fit (1 GiB margin): the 88 GB model does not fit the 48 GB GPU, so llama.cpp's --fit kept the layers on the GPU and placed part of the MoE expert weights in system RAM on the CPU (loaded with -lm none, so the CPU part is in ordinary RAM, not memory-mapped); the exact split is not logged at the default verbosity; 47488 MiB of VRAM in use after load, 8 generation threads, 16 batch threads, flash attention on, 8 server slots of 4096 tokens each, context 32768. GGUF unsloth/Qwen3-235B-A22B-Instruct-2507-GGUF @ 437d6915c5c512869e4cf8b16b840bbe3fb172bc, file Qwen3-235B-A22B-Instruct-2507-UD-Q2_K_XL-00001-of-00002.gguf. Each request: LocalMaxxing's canonical reasoning-v1 prompt with a unique nonce, 256 output tokens. One discarded warm-up batch, then 3 measured batches: 16, 22.4, 15 tok/s; this is the median. Not eligible for the Verified badge, which requires batch 1.
KASA b65-vm: 1 x Intel Arc Pro B65 32 GB (PCIe passthrough, xe driver, Ubuntu 26.04 kernel 7.0) in a VM with 16 vCPU pinned to 8 cores (16 threads) of an AMD EPYC 7343 and 128 GiB. Engine: XeStrata xe0.1.40.2.2 (MistVVK/XeStrata, the Intel fork of Strata; LGPL-3.0) in the chriscorbell/intel-arc-pro-b70-qwen38-flash-next recipe image built from commit 6aad3506 (local image local/xestrata-qwen38:xe0.1.40.2.2, sha256:6f03dde70323). Weights: ISTA-DASLab/Qwen3.8-Flash-Next-GSQ-RCO-GGUF IQ3_S @ ed59f920; the GPU holds the dense layers and a cache of hot experts, the CPU computes the rest of the experts from system RAM at the same time. MTP speculative decoding (4 tokens), int8 KV cache. Settings: SPEC_MIN_P=0.6 VRAM_RESERVE_MIB=1536 MAX_CONTEXT=65536 CONVERSATION_CACHE_MIB=4096. Canonical reasoning-v1, lmx single-shot requests after one discarded warm-up. Measured requests: 52.5, 53.5, 52.9 tok/s; this is the median (tie to the earlier). Campaign ai/vm/b65-vm/2026-10-27-b65-vm-localmaxxing. Prompt/output samples were added from the lmx run file; engineTimingsRaw is XeStrata's own per-request log (decode tokens and tok/s, expert-cache hit rate), because lmx does not capture it for this engine.
KASA b65-vm: 1 x Intel Arc Pro B65 32 GB (PCIe passthrough, xe driver, Ubuntu 26.04 kernel 7.0) in a VM with 16 vCPU pinned to 8 cores (16 threads) of an AMD EPYC 7343 and 128 GiB. Engine: XeStrata xe0.1.40.2.2 (MistVVK/XeStrata, the Intel fork of Strata; LGPL-3.0) in the chriscorbell/intel-arc-pro-b70-qwen38-flash-next recipe image built from commit 6aad3506 (local image local/xestrata-qwen38:xe0.1.40.2.2, sha256:6f03dde70323). Weights: ISTA-DASLab/Qwen3.8-Flash-Next-GSQ-RCO-GGUF IQ3_S @ ed59f920; the GPU holds the dense layers and a cache of hot experts, the CPU computes the rest of the experts from system RAM at the same time. MTP speculative decoding (4 tokens), int8 KV cache. Settings: SPEC_MIN_P=0.58 VRAM_RESERVE_MIB=1024 MAX_CONTEXT=65536 CONVERSATION_CACHE_MIB=2048. Canonical reasoning-v1, lmx single-shot requests after one discarded warm-up. Measured requests: 52.1, 57.9, 58.7 tok/s; this is the median (tie to the earlier). Campaign ai/vm/b65-vm/2026-10-27-b65-vm-localmaxxing. Prompt/output samples were added from the lmx run file; engineTimingsRaw is XeStrata's own per-request log (decode tokens and tok/s, expert-cache hit rate), because lmx does not capture it for this engine.
KASA b65-vm: 1 x Intel Arc Pro B65 32 GB (PCIe passthrough, xe driver, Ubuntu 26.04 kernel 7.0) in a VM with 16 vCPU pinned to 8 cores (16 threads) of an AMD EPYC 7343 and 128 GiB. Engine: XeStrata xe0.1.40.2.2 (MistVVK/XeStrata, the Intel fork of Strata; LGPL-3.0) in the chriscorbell/intel-arc-pro-b70-qwen38-flash-next recipe image built from commit 6aad3506 (local image local/xestrata-qwen38:xe0.1.40.2.2, sha256:6f03dde70323). Weights: ISTA-DASLab/Qwen3.8-Flash-Next-GSQ-RCO-GGUF IQ3_S @ ed59f920; the GPU holds the dense layers and a cache of hot experts, the CPU computes the rest of the experts from system RAM at the same time. MTP speculative decoding (4 tokens), int8 KV cache. Settings: SPEC_MIN_P=0.58 VRAM_RESERVE_MIB=1536 MAX_CONTEXT=65536 CONVERSATION_CACHE_MIB=4096. Canonical reasoning-v1, lmx single-shot requests after one discarded warm-up. Measured requests: 54.2, 55.3, 55.7 tok/s; this is the median (tie to the earlier). Campaign ai/vm/b65-vm/2026-10-27-b65-vm-localmaxxing. Prompt/output samples were added from the lmx run file; engineTimingsRaw is XeStrata's own per-request log (decode tokens and tok/s, expert-cache hit rate), because lmx does not capture it for this engine.
CONCURRENCY TEST, NOT A SINGLE-USER RESULT. 2 requests sent at once (lmx --concurrency 2); 20.5 tok/s is the aggregate decode rate across all 2 streams, about 10.2 tok/s per stream, and the 2329 ms time to first token is under that load. KASA cuda-vm: 1 x NVIDIA RTX A6000 48 GB (PCIe passthrough, driver 615.71.10) in a VM with 16 vCPU pinned to 8 of the 16 cores (16 threads) of an AMD EPYC 7343 (another VM holds the other 8 cores) and 128 GiB. llama.cpp llama-server b11889 (CUDA image ghcr.io/ggml-org/llama.cpp), placement by llama.cpp's default --fit (1 GiB margin): the 88 GB model does not fit the 48 GB GPU, so llama.cpp's --fit kept the layers on the GPU and placed part of the MoE expert weights in system RAM on the CPU (loaded with -lm none, so the CPU part is in ordinary RAM, not memory-mapped); the exact split is not logged at the default verbosity; 47488 MiB of VRAM in use after load, 8 generation threads, 16 batch threads, flash attention on, 8 server slots of 4096 tokens each, context 32768. GGUF unsloth/Qwen3-235B-A22B-Instruct-2507-GGUF @ 437d6915c5c512869e4cf8b16b840bbe3fb172bc, file Qwen3-235B-A22B-Instruct-2507-UD-Q2_K_XL-00001-of-00002.gguf. Each request: LocalMaxxing's canonical reasoning-v1 prompt with a unique nonce, 256 output tokens. One discarded warm-up batch, then 3 measured batches: 18.7, 20.5, 22 tok/s; this is the median. Not eligible for the Verified badge, which requires batch 1. Selection rule applied to every run in this round: the median of the 2 measured requests whose lmx promptTokens is a whole number; the others report a fractional median, which LocalMaxxing refuses.
KASA b65-vm: 1 x Intel Arc Pro B65 32 GB (PCIe passthrough, xe driver, Ubuntu 26.04 kernel 7.0) in a VM with 16 vCPU pinned to 8 cores (16 threads) of an AMD EPYC 7343 and 128 GiB. Engine: XeStrata xe0.1.40.2.2 (MistVVK/XeStrata, the Intel fork of Strata; LGPL-3.0) in the chriscorbell/intel-arc-pro-b70-qwen38-flash-next recipe image built from commit 6aad3506 (local image local/xestrata-qwen38:xe0.1.40.2.2, sha256:6f03dde70323). Weights: ISTA-DASLab/Qwen3.8-Flash-Next-GSQ-RCO-GGUF IQ3_S @ ed59f920; the GPU holds the dense layers and a cache of hot experts, the CPU computes the rest of the experts from system RAM at the same time. MTP speculative decoding (4 tokens), int8 KV cache. Settings: SPEC_MIN_P=0.62 VRAM_RESERVE_MIB=1280 MAX_CONTEXT=65536 CONVERSATION_CACHE_MIB=2048. Canonical reasoning-v1, lmx single-shot requests after one discarded warm-up. Measured requests: 54.4, 61.1, 58.2 tok/s; this is the median (tie to the earlier). Campaign ai/vm/b65-vm/2026-10-27-b65-vm-localmaxxing. Prompt/output samples were added from the lmx run file; engineTimingsRaw is XeStrata's own per-request log (decode tokens and tok/s, expert-cache hit rate), because lmx does not capture it for this engine.
KASA cuda-vm: 1 x NVIDIA RTX A6000 48 GB (PCIe passthrough, driver 615.71.10) in a VM with 16 vCPU pinned to 8 of the 16 cores (16 threads) of an AMD EPYC 7343 (another VM holds the other 8 cores) and 128 GiB. llama.cpp llama-server b11889 (CUDA image ghcr.io/ggml-org/llama.cpp), placement by llama.cpp's default --fit (1 GiB margin): the 88 GB model does not fit the 48 GB GPU, so llama.cpp's --fit kept the layers on the GPU and placed part of the MoE expert weights in system RAM on the CPU (loaded with -lm none, so the CPU part is in ordinary RAM, not memory-mapped); the exact split is not logged at the default verbosity; 47488 MiB of VRAM in use after load, 8 generation threads, 16 batch threads, flash attention on, 8 server slots of 4096 tokens each, context 32768. GGUF unsloth/Qwen3-235B-A22B-Instruct-2507-GGUF @ 437d6915c5c512869e4cf8b16b840bbe3fb172bc, file Qwen3-235B-A22B-Instruct-2507-UD-Q2_K_XL-00001-of-00002.gguf. Canonical reasoning-v1, lmx single-shot requests after one discarded warm-up. Measured requests: 14.1, 14.9, 12.8 tok/s; this is the median (tie to the earlier). Prompt and output samples were added from the lmx run file, engineTimingsRaw from exact llama-server /metrics deltas around the request.
KASA b65-vm: 1 x Intel Arc Pro B65 32 GB (PCIe passthrough, xe driver, Ubuntu 26.04 kernel 7.0) in a VM with 16 vCPU pinned to 8 cores (16 threads) of an AMD EPYC 7343 and 128 GiB. Engine: XeStrata xe0.1.40.2.2 (MistVVK/XeStrata, the Intel fork of Strata; LGPL-3.0) in the chriscorbell/intel-arc-pro-b70-qwen38-flash-next recipe image built from commit 6aad3506 (local image local/xestrata-qwen38:xe0.1.40.2.2, sha256:6f03dde70323). Weights: ISTA-DASLab/Qwen3.8-Flash-Next-GSQ-RCO-GGUF IQ3_S @ ed59f920; the GPU holds the dense layers and a cache of hot experts, the CPU computes the rest of the experts from system RAM at the same time. MTP speculative decoding (4 tokens), int8 KV cache. Settings: SPEC_MIN_P=0.58 VRAM_RESERVE_MIB=1024 MAX_CONTEXT=131072 CONVERSATION_CACHE_MIB=4096. Canonical reasoning-v1, lmx single-shot requests after one discarded warm-up. Measured requests: 54, 57.2, 59.5 tok/s; this is the median (tie to the earlier). Campaign ai/vm/b65-vm/2026-10-27-b65-vm-localmaxxing. Prompt/output samples were added from the lmx run file; engineTimingsRaw is XeStrata's own per-request log (decode tokens and tok/s, expert-cache hit rate), because lmx does not capture it for this engine.
KASA b65-vm: 1 x Intel Arc Pro B65 32 GB (PCIe passthrough, xe driver, Ubuntu 26.04 kernel 7.0) in a VM with 16 vCPU pinned to 8 cores (16 threads) of an AMD EPYC 7343 and 128 GiB. Engine: XeStrata xe0.1.40.2.2 (MistVVK/XeStrata, the Intel fork of Strata; LGPL-3.0) in the chriscorbell/intel-arc-pro-b70-qwen38-flash-next recipe image built from commit 6aad3506 (local image local/xestrata-qwen38:xe0.1.40.2.2, sha256:6f03dde70323). Weights: ISTA-DASLab/Qwen3.8-Flash-Next-GSQ-RCO-GGUF IQ3_S @ ed59f920; the GPU holds the dense layers and a cache of hot experts, the CPU computes the rest of the experts from system RAM at the same time. MTP speculative decoding (4 tokens), int8 KV cache. Settings: SPEC_MIN_P=0.58 VRAM_RESERVE_MIB=1024 MAX_CONTEXT=131072 CONVERSATION_CACHE_MIB=2048. Canonical reasoning-v1, lmx single-shot requests after one discarded warm-up. Measured requests: 40, 54, 55.8 tok/s; this is the median (tie to the earlier). Campaign ai/vm/b65-vm/2026-10-27-b65-vm-localmaxxing. Prompt/output samples were added from the lmx run file; engineTimingsRaw is XeStrata's own per-request log (decode tokens and tok/s, expert-cache hit rate), because lmx does not capture it for this engine.
KASA b65-vm: 1 x Intel Arc Pro B65 32 GB (PCIe passthrough, xe driver, Ubuntu 26.04 kernel 7.0) in a VM with 16 vCPU pinned to 8 cores (16 threads) of an AMD EPYC 7343 and 128 GiB. Engine: XeStrata xe0.1.40.2.2 (MistVVK/XeStrata, the Intel fork of Strata; LGPL-3.0) in the chriscorbell/intel-arc-pro-b70-qwen38-flash-next recipe image built from commit 6aad3506 (local image local/xestrata-qwen38:xe0.1.40.2.2, sha256:6f03dde70323). Weights: ISTA-DASLab/Qwen3.8-Flash-Next-GSQ-RCO-GGUF IQ3_S @ ed59f920; the GPU holds the dense layers and a cache of hot experts, the CPU computes the rest of the experts from system RAM at the same time. MTP speculative decoding (4 tokens), int8 KV cache. Settings: SPEC_MIN_P=0.58 VRAM_RESERVE_MIB=1536 MAX_CONTEXT=65536 CONVERSATION_CACHE_MIB=2048. Canonical reasoning-v1, lmx single-shot requests after one discarded warm-up. Measured requests: 51, 57.6, 60.3 tok/s; this is the median (tie to the earlier). Campaign ai/vm/b65-vm/2026-10-27-b65-vm-localmaxxing. Prompt/output samples were added from the lmx run file; engineTimingsRaw is XeStrata's own per-request log (decode tokens and tok/s, expert-cache hit rate), because lmx does not capture it for this engine.
KASA b65-vm: 1 x Intel Arc Pro B65 32 GB (PCIe passthrough, xe driver, Ubuntu 26.04 kernel 7.0) in a VM with 16 vCPU pinned to 8 cores (16 threads) of an AMD EPYC 7343 and 128 GiB. Engine: XeStrata xe0.1.40.2.2 (MistVVK/XeStrata, the Intel fork of Strata; LGPL-3.0) in the chriscorbell/intel-arc-pro-b70-qwen38-flash-next recipe image built from commit 6aad3506 (local image local/xestrata-qwen38:xe0.1.40.2.2, sha256:6f03dde70323). Weights: ISTA-DASLab/Qwen3.8-Flash-Next-GSQ-RCO-GGUF IQ3_S @ ed59f920; the GPU holds the dense layers and a cache of hot experts, the CPU computes the rest of the experts from system RAM at the same time. MTP speculative decoding (4 tokens), int8 KV cache. Settings: SPEC_MIN_P=0.6 VRAM_RESERVE_MIB=1536 MAX_CONTEXT=131072 CONVERSATION_CACHE_MIB=4096. Canonical reasoning-v1, lmx single-shot requests after one discarded warm-up. Measured requests: 51.8, 54, 55.4 tok/s; this is the median (tie to the earlier). Campaign ai/vm/b65-vm/2026-10-27-b65-vm-localmaxxing. Prompt/output samples were added from the lmx run file; engineTimingsRaw is XeStrata's own per-request log (decode tokens and tok/s, expert-cache hit rate), because lmx does not capture it for this engine.
KASA b65-vm: 1 x Intel Arc Pro B65 32 GB (PCIe passthrough, xe driver, Ubuntu 26.04 kernel 7.0) in a VM with 16 vCPU pinned to 8 cores (16 threads) of an AMD EPYC 7343 and 128 GiB. Engine: XeStrata xe0.1.40.2.2 (MistVVK/XeStrata, the Intel fork of Strata; LGPL-3.0) in the chriscorbell/intel-arc-pro-b70-qwen38-flash-next recipe image built from commit 6aad3506 (local image local/xestrata-qwen38:xe0.1.40.2.2, sha256:6f03dde70323). Weights: ISTA-DASLab/Qwen3.8-Flash-Next-GSQ-RCO-GGUF IQ3_S @ ed59f920; the GPU holds the dense layers and a cache of hot experts, the CPU computes the rest of the experts from system RAM at the same time. MTP speculative decoding (4 tokens), int8 KV cache. Settings: SPEC_MIN_P=0.58 VRAM_RESERVE_MIB=1536 MAX_CONTEXT=131072 CONVERSATION_CACHE_MIB=2048. Canonical reasoning-v1, lmx single-shot requests after one discarded warm-up. Measured requests: 50.7, 56.8, 54.8 tok/s; this is the median (tie to the earlier). Campaign ai/vm/b65-vm/2026-10-27-b65-vm-localmaxxing. Prompt/output samples were added from the lmx run file; engineTimingsRaw is XeStrata's own per-request log (decode tokens and tok/s, expert-cache hit rate), because lmx does not capture it for this engine.
KASA b65-vm: 1 x Intel Arc Pro B65 32 GB (PCIe passthrough, xe driver, Ubuntu 26.04 kernel 7.0) in a VM with 16 vCPU pinned to 8 cores (16 threads) of an AMD EPYC 7343 and 128 GiB. Engine: XeStrata xe0.1.40.2.2 (MistVVK/XeStrata, the Intel fork of Strata; LGPL-3.0) in the chriscorbell/intel-arc-pro-b70-qwen38-flash-next recipe image built from commit 6aad3506 (local image local/xestrata-qwen38:xe0.1.40.2.2, sha256:6f03dde70323). Weights: ISTA-DASLab/Qwen3.8-Flash-Next-GSQ-RCO-GGUF IQ3_S @ ed59f920; the GPU holds the dense layers and a cache of hot experts, the CPU computes the rest of the experts from system RAM at the same time. MTP speculative decoding (4 tokens), int8 KV cache. Settings: SPEC_MIN_P=0.58 VRAM_RESERVE_MIB=1280 MAX_CONTEXT=131072 CONVERSATION_CACHE_MIB=4096. Canonical reasoning-v1, lmx single-shot requests after one discarded warm-up. Measured requests: 50.5, 57.6, 59.4 tok/s; this is the median (tie to the earlier). Campaign ai/vm/b65-vm/2026-10-27-b65-vm-localmaxxing. Prompt/output samples were added from the lmx run file; engineTimingsRaw is XeStrata's own per-request log (decode tokens and tok/s, expert-cache hit rate), because lmx does not capture it for this engine.
KASA b65-vm: 1 x Intel Arc Pro B65 32 GB (PCIe passthrough, xe driver, Ubuntu 26.04 kernel 7.0) in a VM with 16 vCPU pinned to 8 cores (16 threads) of an AMD EPYC 7343 and 128 GiB. Engine: XeStrata xe0.1.40.2.2 (MistVVK/XeStrata, the Intel fork of Strata; LGPL-3.0) in the chriscorbell/intel-arc-pro-b70-qwen38-flash-next recipe image built from commit 6aad3506 (local image local/xestrata-qwen38:xe0.1.40.2.2, sha256:6f03dde70323). Weights: ISTA-DASLab/Qwen3.8-Flash-Next-GSQ-RCO-GGUF IQ3_S @ ed59f920; the GPU holds the dense layers and a cache of hot experts, the CPU computes the rest of the experts from system RAM at the same time. MTP speculative decoding (4 tokens), int8 KV cache. Settings: SPEC_MIN_P=0.62 VRAM_RESERVE_MIB=1536 MAX_CONTEXT=131072 CONVERSATION_CACHE_MIB=2048. Canonical reasoning-v1, lmx single-shot requests after one discarded warm-up. Measured requests: 54.8, 56, 56.3 tok/s; this is the median (tie to the earlier). Campaign ai/vm/b65-vm/2026-10-27-b65-vm-localmaxxing. Prompt/output samples were added from the lmx run file; engineTimingsRaw is XeStrata's own per-request log (decode tokens and tok/s, expert-cache hit rate), because lmx does not capture it for this engine.
KASA b65-vm: 1 x Intel Arc Pro B65 32 GB (PCIe passthrough, xe driver, Ubuntu 26.04 kernel 7.0) in a VM with 16 vCPU pinned to 8 cores (16 threads) of an AMD EPYC 7343 and 128 GiB. Engine: XeStrata xe0.1.40.2.2 (MistVVK/XeStrata, the Intel fork of Strata; LGPL-3.0) in the chriscorbell/intel-arc-pro-b70-qwen38-flash-next recipe image built from commit 6aad3506 (local image local/xestrata-qwen38:xe0.1.40.2.2, sha256:6f03dde70323). Weights: ISTA-DASLab/Qwen3.8-Flash-Next-GSQ-RCO-GGUF IQ3_S @ ed59f920; the GPU holds the dense layers and a cache of hot experts, the CPU computes the rest of the experts from system RAM at the same time. MTP speculative decoding (4 tokens), int8 KV cache. Settings: SPEC_MIN_P=0.6 VRAM_RESERVE_MIB=1280 MAX_CONTEXT=131072 CONVERSATION_CACHE_MIB=2048. Canonical reasoning-v1, lmx single-shot requests after one discarded warm-up. Measured requests: 53.3, 58, 55 tok/s; this is the median (tie to the earlier). Campaign ai/vm/b65-vm/2026-10-27-b65-vm-localmaxxing. Prompt/output samples were added from the lmx run file; engineTimingsRaw is XeStrata's own per-request log (decode tokens and tok/s, expert-cache hit rate), because lmx does not capture it for this engine.
KASA b65-vm: 1 x Intel Arc Pro B65 32 GB (PCIe passthrough, xe driver, Ubuntu 26.04 kernel 7.0) in a VM with 16 vCPU pinned to 8 cores (16 threads) of an AMD EPYC 7343 and 128 GiB. Engine: XeStrata xe0.1.40.2.2 (MistVVK/XeStrata, the Intel fork of Strata; LGPL-3.0) in the chriscorbell/intel-arc-pro-b70-qwen38-flash-next recipe image built from commit 6aad3506 (local image local/xestrata-qwen38:xe0.1.40.2.2, sha256:6f03dde70323). Weights: ISTA-DASLab/Qwen3.8-Flash-Next-GSQ-RCO-GGUF IQ3_S @ ed59f920; the GPU holds the dense layers and a cache of hot experts, the CPU computes the rest of the experts from system RAM at the same time. MTP speculative decoding (4 tokens), int8 KV cache. Settings: SPEC_MIN_P=0.62 VRAM_RESERVE_MIB=1280 MAX_CONTEXT=131072 CONVERSATION_CACHE_MIB=4096. Canonical reasoning-v1, lmx single-shot requests after one discarded warm-up. Measured requests: 51.7, 59.6, 56.3 tok/s; this is the median (tie to the earlier). Campaign ai/vm/b65-vm/2026-10-27-b65-vm-localmaxxing. Prompt/output samples were added from the lmx run file; engineTimingsRaw is XeStrata's own per-request log (decode tokens and tok/s, expert-cache hit rate), because lmx does not capture it for this engine.
KASA b65-vm: 1 x Intel Arc Pro B65 32 GB (PCIe passthrough, xe driver, Ubuntu 26.04 kernel 7.0) in a VM with 16 vCPU pinned to 8 cores (16 threads) of an AMD EPYC 7343 and 128 GiB. Engine: XeStrata xe0.1.40.2.2 (MistVVK/XeStrata, the Intel fork of Strata; LGPL-3.0) in the chriscorbell/intel-arc-pro-b70-qwen38-flash-next recipe image built from commit 6aad3506 (local image local/xestrata-qwen38:xe0.1.40.2.2, sha256:6f03dde70323). Weights: ISTA-DASLab/Qwen3.8-Flash-Next-GSQ-RCO-GGUF IQ3_S @ ed59f920; the GPU holds the dense layers and a cache of hot experts, the CPU computes the rest of the experts from system RAM at the same time. MTP speculative decoding (4 tokens), int8 KV cache. Settings: SPEC_MIN_P=0.62 VRAM_RESERVE_MIB=1536 MAX_CONTEXT=131072 CONVERSATION_CACHE_MIB=4096. Canonical reasoning-v1, lmx single-shot requests after one discarded warm-up. Measured requests: 54.7, 57.5, 58.1 tok/s; this is the median (tie to the earlier). Campaign ai/vm/b65-vm/2026-10-27-b65-vm-localmaxxing. Prompt/output samples were added from the lmx run file; engineTimingsRaw is XeStrata's own per-request log (decode tokens and tok/s, expert-cache hit rate), because lmx does not capture it for this engine.
KASA b65-vm: 1 x Intel Arc Pro B65 32 GB (PCIe passthrough, xe driver, Ubuntu 26.04 kernel 7.0) in a VM with 16 vCPU pinned to 8 cores (16 threads) of an AMD EPYC 7343 and 128 GiB. Engine: XeStrata xe0.1.40.2.2 (MistVVK/XeStrata, the Intel fork of Strata; LGPL-3.0) in the chriscorbell/intel-arc-pro-b70-qwen38-flash-next recipe image built from commit 6aad3506 (local image local/xestrata-qwen38:xe0.1.40.2.2, sha256:6f03dde70323). Weights: ISTA-DASLab/Qwen3.8-Flash-Next-GSQ-RCO-GGUF IQ3_S @ ed59f920; the GPU holds the dense layers and a cache of hot experts, the CPU computes the rest of the experts from system RAM at the same time. MTP speculative decoding (4 tokens), int8 KV cache. Settings: SPEC_MIN_P=0.6 VRAM_RESERVE_MIB=1280 MAX_CONTEXT=131072 CONVERSATION_CACHE_MIB=4096. Canonical reasoning-v1, lmx single-shot requests after one discarded warm-up. Measured requests: 51.8, 58.4, 56.5 tok/s; this is the median (tie to the earlier). Campaign ai/vm/b65-vm/2026-10-27-b65-vm-localmaxxing. Prompt/output samples were added from the lmx run file; engineTimingsRaw is XeStrata's own per-request log (decode tokens and tok/s, expert-cache hit rate), because lmx does not capture it for this engine.
CONCURRENCY TEST, NOT A SINGLE-USER RESULT. 6 requests sent at once (lmx --concurrency 6); 5.1 tok/s is the aggregate decode rate across all 6 streams, about 0.8 tok/s per stream, and the 12149 ms time to first token is under that load. KASA cuda-vm: 1 x NVIDIA RTX A6000 48 GB (PCIe passthrough, driver 615.71.10) in a VM with 16 vCPU pinned to 8 of the 16 cores (16 threads) of an AMD EPYC 7343 (another VM holds the other 8 cores) and 128 GiB. llama.cpp llama-server b11889 (CUDA image ghcr.io/ggml-org/llama.cpp), placement by llama.cpp's default --fit (1 GiB margin): the 122 GB model does not fit the 48 GB GPU, so llama.cpp's --fit kept the layers on the GPU and placed part of the MoE expert weights in system RAM on the CPU (loaded with -lm none, so the CPU part is in ordinary RAM, not memory-mapped); the exact split is not logged at the default verbosity; 47658 MiB of VRAM in use after load, 8 generation threads, 16 batch threads, flash attention on, 8 server slots of 4096 tokens each, context 32768. GGUF unsloth/GLM-4.7-GGUF @ 70deda3fcbe622671ff2111d7c0c7392f618e3b0, file GLM-4.7-UD-IQ2_M-00001-of-00003.gguf. Each request: LocalMaxxing's canonical reasoning-v1 prompt with a unique nonce, 256 output tokens. One discarded warm-up batch, then 3 measured batches: 3.4, 5.1, 3.8 tok/s; this is the median. Not eligible for the Verified badge, which requires batch 1. Selection rule applied to every run in this round: the median of the 1 measured requests whose lmx promptTokens is a whole number; the others report a fractional median, which LocalMaxxing refuses.
KASA b65-vm: 1 x Intel Arc Pro B65 32 GB (PCIe passthrough, xe driver, Ubuntu 26.04 kernel 7.0) in a VM with 16 vCPU pinned to 8 cores (16 threads) of an AMD EPYC 7343 and 128 GiB. Engine: XeStrata xe0.1.40.2.2 (MistVVK/XeStrata, the Intel fork of Strata; LGPL-3.0) in the chriscorbell/intel-arc-pro-b70-qwen38-flash-next recipe image built from commit 6aad3506 (local image local/xestrata-qwen38:xe0.1.40.2.2, sha256:6f03dde70323). Weights: ISTA-DASLab/Qwen3.8-Flash-Next-GSQ-RCO-GGUF IQ3_S @ ed59f920; the GPU holds the dense layers and a cache of hot experts, the CPU computes the rest of the experts from system RAM at the same time. MTP speculative decoding (4 tokens), int8 KV cache. Settings: SPEC_MIN_P=0.58 VRAM_RESERVE_MIB=1280 MAX_CONTEXT=65536 CONVERSATION_CACHE_MIB=4096. Canonical reasoning-v1, lmx single-shot requests after one discarded warm-up. Measured requests: 46.4, 51.2, 53.2 tok/s; this is the median (tie to the earlier). Campaign ai/vm/b65-vm/2026-10-27-b65-vm-localmaxxing. Prompt/output samples were added from the lmx run file; engineTimingsRaw is XeStrata's own per-request log (decode tokens and tok/s, expert-cache hit rate), because lmx does not capture it for this engine.
KASA b65-vm: 1 x Intel Arc Pro B65 32 GB (PCIe passthrough, xe driver, Ubuntu 26.04 kernel 7.0) in a VM with 16 vCPU pinned to 8 cores (16 threads) of an AMD EPYC 7343 and 128 GiB. Engine: XeStrata xe0.1.40.2.2 (MistVVK/XeStrata, the Intel fork of Strata; LGPL-3.0) in the chriscorbell/intel-arc-pro-b70-qwen38-flash-next recipe image built from commit 6aad3506 (local image local/xestrata-qwen38:xe0.1.40.2.2, sha256:6f03dde70323). Weights: ISTA-DASLab/Qwen3.8-Flash-Next-GSQ-RCO-GGUF IQ3_S @ ed59f920; the GPU holds the dense layers and a cache of hot experts, the CPU computes the rest of the experts from system RAM at the same time. MTP speculative decoding (4 tokens), int8 KV cache. Settings: SPEC_MIN_P=0.62 VRAM_RESERVE_MIB=1024 MAX_CONTEXT=131072 CONVERSATION_CACHE_MIB=2048. Canonical reasoning-v1, lmx single-shot requests after one discarded warm-up. Measured requests: 48.2, 54.4, 57.5 tok/s; this is the median (tie to the earlier). Campaign ai/vm/b65-vm/2026-10-27-b65-vm-localmaxxing. Prompt/output samples were added from the lmx run file; engineTimingsRaw is XeStrata's own per-request log (decode tokens and tok/s, expert-cache hit rate), because lmx does not capture it for this engine.
KASA b65-vm: 1 x Intel Arc Pro B65 32 GB (PCIe passthrough, xe driver, Ubuntu 26.04 kernel 7.0) in a VM with 16 vCPU pinned to 8 cores (16 threads) of an AMD EPYC 7343 and 128 GiB. Engine: XeStrata xe0.1.40.2.2 (MistVVK/XeStrata, the Intel fork of Strata; LGPL-3.0) in the chriscorbell/intel-arc-pro-b70-qwen38-flash-next recipe image built from commit 6aad3506 (local image local/xestrata-qwen38:xe0.1.40.2.2, sha256:6f03dde70323). Weights: ISTA-DASLab/Qwen3.8-Flash-Next-GSQ-RCO-GGUF IQ3_S @ ed59f920; the GPU holds the dense layers and a cache of hot experts, the CPU computes the rest of the experts from system RAM at the same time. MTP speculative decoding (4 tokens), int8 KV cache. Settings: SPEC_MIN_P=0.62 VRAM_RESERVE_MIB=1536 MAX_CONTEXT=65536 CONVERSATION_CACHE_MIB=4096. Canonical reasoning-v1, lmx single-shot requests after one discarded warm-up. Measured requests: 49.3, 56.6, 55.6 tok/s; this is the median (tie to the earlier). Campaign ai/vm/b65-vm/2026-10-27-b65-vm-localmaxxing. Prompt/output samples were added from the lmx run file; engineTimingsRaw is XeStrata's own per-request log (decode tokens and tok/s, expert-cache hit rate), because lmx does not capture it for this engine.
KASA b65-vm: 1 x Intel Arc Pro B65 32 GB (PCIe passthrough, xe driver, Ubuntu 26.04 kernel 7.0) in a VM with 16 vCPU pinned to 8 cores (16 threads) of an AMD EPYC 7343 and 128 GiB. Engine: XeStrata xe0.1.40.2.2 (MistVVK/XeStrata, the Intel fork of Strata; LGPL-3.0) in the chriscorbell/intel-arc-pro-b70-qwen38-flash-next recipe image built from commit 6aad3506 (local image local/xestrata-qwen38:xe0.1.40.2.2, sha256:6f03dde70323). Weights: ISTA-DASLab/Qwen3.8-Flash-Next-GSQ-RCO-GGUF IQ3_S @ ed59f920; the GPU holds the dense layers and a cache of hot experts, the CPU computes the rest of the experts from system RAM at the same time. MTP speculative decoding (4 tokens), int8 KV cache. Settings: SPEC_MIN_P=0.62 VRAM_RESERVE_MIB=1024 MAX_CONTEXT=131072 CONVERSATION_CACHE_MIB=4096. Canonical reasoning-v1, lmx single-shot requests after one discarded warm-up. Measured requests: 53.4, 57.5, 60.3 tok/s; this is the median (tie to the earlier). Campaign ai/vm/b65-vm/2026-10-27-b65-vm-localmaxxing. Prompt/output samples were added from the lmx run file; engineTimingsRaw is XeStrata's own per-request log (decode tokens and tok/s, expert-cache hit rate), because lmx does not capture it for this engine.
KASA b65-vm: 1 x Intel Arc Pro B65 32 GB (PCIe passthrough, xe driver, Ubuntu 26.04 kernel 7.0) in a VM with 16 vCPU pinned to 8 cores (16 threads) of an AMD EPYC 7343 and 128 GiB. Engine: XeStrata xe0.1.40.2.2 (MistVVK/XeStrata, the Intel fork of Strata; LGPL-3.0) in the chriscorbell/intel-arc-pro-b70-qwen38-flash-next recipe image built from commit 6aad3506 (local image local/xestrata-qwen38:xe0.1.40.2.2, sha256:6f03dde70323). Weights: ISTA-DASLab/Qwen3.8-Flash-Next-GSQ-RCO-GGUF IQ3_S @ ed59f920; the GPU holds the dense layers and a cache of hot experts, the CPU computes the rest of the experts from system RAM at the same time. MTP speculative decoding (4 tokens), int8 KV cache. Settings: SPEC_MIN_P=0.58 VRAM_RESERVE_MIB=1280 MAX_CONTEXT=65536 CONVERSATION_CACHE_MIB=2048. Canonical reasoning-v1, lmx single-shot requests after one discarded warm-up. Measured requests: 50.4, 55.4, 60.1 tok/s; this is the median (tie to the earlier). Campaign ai/vm/b65-vm/2026-10-27-b65-vm-localmaxxing. Prompt/output samples were added from the lmx run file; engineTimingsRaw is XeStrata's own per-request log (decode tokens and tok/s, expert-cache hit rate), because lmx does not capture it for this engine.
KASA b65-vm: 1 x Intel Arc Pro B65 32 GB (PCIe passthrough, xe driver, Ubuntu 26.04 kernel 7.0) in a VM with 16 vCPU pinned to 8 cores (16 threads) of an AMD EPYC 7343 and 128 GiB. Engine: XeStrata xe0.1.40.2.2 (MistVVK/XeStrata, the Intel fork of Strata; LGPL-3.0) in the chriscorbell/intel-arc-pro-b70-qwen38-flash-next recipe image built from commit 6aad3506 (local image local/xestrata-qwen38:xe0.1.40.2.2, sha256:6f03dde70323). Weights: ISTA-DASLab/Qwen3.8-Flash-Next-GSQ-RCO-GGUF IQ3_S @ ed59f920; the GPU holds the dense layers and a cache of hot experts, the CPU computes the rest of the experts from system RAM at the same time. MTP speculative decoding (4 tokens), int8 KV cache. Settings: SPEC_MIN_P=0.6 VRAM_RESERVE_MIB=1280 MAX_CONTEXT=65536 CONVERSATION_CACHE_MIB=4096. Canonical reasoning-v1, lmx single-shot requests after one discarded warm-up. Measured requests: 53, 57.8, 55.1 tok/s; this is the median (tie to the earlier). Campaign ai/vm/b65-vm/2026-10-27-b65-vm-localmaxxing. Prompt/output samples were added from the lmx run file; engineTimingsRaw is XeStrata's own per-request log (decode tokens and tok/s, expert-cache hit rate), because lmx does not capture it for this engine.
