入門指南排行榜Decode calculator模型OrganizationsReportsMatrix runs硬體基準評測市集租用ProAPI文件
語言
Actual Computer — Every computer, one endpoint

Matrix runs

A matrix run measures prompt reading (prefill) and generation (decode) speed in one go, across several context lengths and numbers of simultaneous requests. Run one with lmx speed-test matrix run.

All runs 740

Newest first. Page 12 of 30.
  • Qwen3.8-Flash-NextRTX PRO 6000 Blackwell Server Edition 96GB ×4 · vllm 0.1.dev1+g50a061f79 · BF166 contexts × 8 concurrency levels
    Peak decode
    1,574.4 tok/s
    Peak prefill
    14.4K tok/s
    by @keennay
  • Qwen3.8-Flash-NextRTX PRO 6000 Blackwell Server Edition 96GB ×8 · vllm 0.1.1.dev5+ge77daef89 · BF166 contexts × 8 concurrency levels
    Peak decode
    3,084.6 tok/s
    Peak prefill
    14.6K tok/s
    by @keennay
  • DeepSeek-V4.1-Flash-NVFP4RTX PRO 6000 Blackwell Server Edition 96GB ×8 · vllm 0.30.1rc1.dev147+g02ee13be2 · NVFP4+MXFP4+FP88 contexts × 8 concurrency levels
    Peak decode
    590.6 tok/s
    Peak prefill
    9,885 tok/s
    by @keennay
  • DeepSeek-V4.1-Flash-NVFP4RTX PRO 6000 Blackwell Server Edition 96GB ×8 · vllm 0.30.1rc1.dev147+g02ee13be2 · NVFP4+MXFP4+FP86 contexts × 8 concurrency levels
    Peak decode
    597.2 tok/s
    Peak prefill
    10.1K tok/s
    by @keennay
  • DeepSeek-V4.1-FlashRTX PRO 6000 Blackwell Server Edition 96GB ×8 · vllm 0.30.1rc1.dev147+g02ee13be2 · MXFP4+FP88 contexts × 8 concurrency levels
    Peak decode
    3,066.1 tok/s
    Peak prefill
    10.6K tok/s
    by @keennay
  • Qwen3.8-Flash-Next-NVFP4RTX PRO 6000 Blackwell Server Edition 96GB ×4 · vllm 0.28.1rc1.dev562+g9c2d21046 · NVFP4+FP86 contexts × 8 concurrency levels
    Peak decode
    4,428.5 tok/s
    Peak prefill
    16.9K tok/s
    by @keennay
  • GLM-5.3-Flash-NVFP4RTX PRO 6000 Blackwell Server Edition 96GB ×4 · sglang 0.5.6.post3.dev10352+g28379a7c5 · NVFP48 contexts × 8 concurrency levels
    Peak decode
    1,870.1 tok/s
    Peak prefill
    12K tok/s
    by @keennay
  • GLM-5.3-NVFP4RTX PRO 6000 Blackwell Server Edition 96GB ×8 · sglang 0.5.6.post3.dev9606+g26fd7fdaa · NVFP4+FP86 contexts × 8 concurrency levels
    Peak decode
    476.3 tok/s
    Peak prefill
    4,309 tok/s
    by @keennay
  • DeepSeek-V4.1-Flash-NVFP4H200 SXM 141GB ×8 · vllm 0.1.1.dev5+ge77daef89 · NVFP4+MXFP4+FP88 contexts × 8 concurrency levels
    Peak decode
    5,533.8 tok/s
    Peak prefill
    21.6K tok/s
    by @keennay
  • DeepSeek-V4.1-Flash-NVFP4H200 SXM 141GB ×4 · vllm 0.1.1.dev5+ge77daef89 · NVFP4+MXFP4+FP88 contexts × 8 concurrency levels
    Peak decode
    4,472.3 tok/s
    Peak prefill
    16.9K tok/s
    by @keennay
  • GLM-5.3-NVFP4H200 SXM 141GB ×8 · sglang 0.5.6.post3.dev9606+g26fd7fdaa · NVFP4+FP87 contexts × 8 concurrency levels
    Peak decode
    1,999.7 tok/s
    Peak prefill
    8,067 tok/s
    by @keennay
  • MiMo-V2.6-Pro-RLH200 SXM 141GB ×8 · vllm 0.29.1rc1.dev533+g1ea7c63f4 · MXFP4+FP88 contexts × 8 concurrency levels
    Peak decode
    3,636.7 tok/s
    Peak prefill
    16.1K tok/s
    by @keennay
  • MiMo-V2.6-Pro-RLH200 SXM 141GB ×8 · vllm 0.29.1rc1.dev533+g1ea7c63f4 · MXFP4+FP88 contexts × 8 concurrency levels
    Peak decode
    2,889.4 tok/s
    Peak prefill
    16.1K tok/s
    by @keennay
  • MiMo-V2.6-Flash-RLH200 NVL 141GB ×2 · vllm 0.29.1rc1.dev533+g1ea7c63f4 · MXFP4+FP88 contexts × 8 concurrency levels
    Peak decode
    2,986.6 tok/s
    Peak prefill
    12.6K tok/s
    by @keennay
  • MiMo-V2.6-Flash-RLH200 NVL 141GB ×2 · vllm 0.29.1rc1.dev533+g1ea7c63f4 · MXFP4+FP88 contexts × 8 concurrency levels
    Peak decode
    1,447.6 tok/s
    Peak prefill
    12.6K tok/s
    by @keennay
  • limite-1b-baseH200 NVL 141GB · vllm 0.26.0 · BF165 contexts × 8 concurrency levels
    Peak decode
    6,879.0 tok/s
    Peak prefill
    103.6K tok/s
    by @keennay
  • limite-1b-base-soupH200 NVL 141GB · vllm 0.26.0 · BF165 contexts × 8 concurrency levels
    Peak decode
    15.3K tok/s
    Peak prefill
    98.3K tok/s
    by @keennay
  • limite-1b-violettoH200 NVL 141GB · vllm 0.26.0 · BF165 contexts × 8 concurrency levels
    Peak decode
    986.2 tok/s
    Peak prefill
    88.9K tok/s
    by @keennay
  • MiMo-V2.6-Distill-Qwen-9BH200 NVL 141GB · sglang 0.5.20 · BF166 contexts × 8 concurrency levels
    Peak burst decode
    —
    Peak prefill
    24.8K tok/s
    by @keennay
  • MiMo-V2.6-Distill-Qwen-9BH200 NVL 141GB · vllm 0.30.0 · BF166 contexts × 8 concurrency levels
    Peak decode
    4,790.4 tok/s
    Peak prefill
    25K tok/s
    by @keennay
  • Muse-Glimmer-30B-NVFP4H200 NVL 141GB · sglang 0.5.20 · NVFP4+FP85 contexts × 8 concurrency levels
    Peak burst decode
    —
    Peak prefill
    6,390 tok/s
    by @keennay
  • Muse-Glimmer-30B-NVFP4H200 NVL 141GB · vllm 0.29.0 · NVFP4+FP85 contexts × 8 concurrency levels
    Peak decode
    1,704.4 tok/s
    Peak prefill
    6,312 tok/s
    by @keennay
  • Qwen3.5-0.8BRTX PRO 6000 Blackwell Server Edition 96GB · sglang 0.5.6.post3.dev9160+g1cf2b8c54 · BF166 contexts × 8 concurrency levels
    Peak decode
    6,829.2 tok/s
    Peak prefill
    82.7K tok/s
    by @keennay
  • Qwen3.5-0.8BRTX PRO 6000 Blackwell Server Edition 96GB · sglang 0.5.6.post3.dev9160+g1cf2b8c54 · BF166 contexts × 8 concurrency levels
    Peak decode
    11.5K tok/s
    Peak prefill
    88.8K tok/s
    by @keennay
  • Qwen3.5-0.8B-BaseRTX PRO 6000 Blackwell Server Edition 96GB · sglang 0.5.6.post3.dev9160+g1cf2b8c54 · BF166 contexts × 8 concurrency levels
    Peak decode
    3,371.7 tok/s
    Peak prefill
    81K tok/s
    by @keennay