./kernelbench

all runs

84 runs · 51 pass · 15 fail · 18 err · sorted by peak_fraction desc

One row per (model, problem) cell. Click any row to open the full transcript viewer — every tool call, every reasoning step, the model's solution.py, the check.log, the result.json. The viewer is the same one we use locally to audit runs, just themed for the site.

peakproblemmodelharnesselapsedrun id
0.60203_paged_attentionclaude-opus-4-7 [max]claude45m20260428_053541_claude_claude-opus-4-7_03_paged_attention
0.53401_fp8_gemmclaude-opus-4-7 [max]claude45m20260428_040539_claude_claude-opus-4-7_01_fp8_gemm
0.49703_paged_attentiongpt-5.5 [xhigh]codex10m20260427_235415_codex_gpt-5.5_03_paged_attention
0.43401_fp8_gemmor/xiaomi/mimo-v2.5-proopencode41m20260428_184159_opencode_openrouter-pinned_xiaomi_mimo-v2.5-pro_01_fp8_gemm
0.43203_paged_attentionkimi-k2.6kimi44m20260428_110049_kimi_kimi-k2.6_03_paged_attention
0.43101_fp8_gemmor/qwen/qwen3.6-plusopencode45m20260429_061331_opencode_openrouter-pinned_qwen_qwen3.6-plus_01_fp8_gemm
0.42901_fp8_gemmor/qwen/qwen3.6-max-previewopencode39m20260428_225859_opencode_openrouter-pinned_qwen_qwen3.6-max-preview_01_fp8_gemm
0.42301_fp8_gemmgpt-5.5 [xhigh]codex16m20260427_230711_codex_gpt-5.5_01_fp8_gemm
0.36304_kahan_softmaxgpt-5.5 [xhigh]codex10m20260428_000416_codex_gpt-5.5_04_kahan_softmax
0.31704_kahan_softmaxclaude-opus-4-7 [max]claude45m20260428_062104_claude_claude-opus-4-7_04_kahan_softmax
0.25106_sonic_moe_swiglugpt-5.5 [xhigh]codex26m20260428_003928_codex_gpt-5.5_06_sonic_moe_swiglu
0.23806_sonic_moe_swigluzai/glm-5.1opencode45m20260428_170757_opencode_zai_glm-5.1_06_sonic_moe_swiglu
0.22007_w4a16_gemmkimi-k2.6kimi45m20260428_134617_kimi_kimi-k2.6_07_w4a16_gemm
0.21106_sonic_moe_swigluor/xiaomi/mimo-v2.5-proopencode45m20260428_213514_opencode_openrouter-pinned_xiaomi_mimo-v2.5-pro_06_sonic_moe_swiglu
0.18407_w4a16_gemmclaude-opus-4-7 [max]claude45m20260428_083917_claude_claude-opus-4-7_07_w4a16_gemm
0.18007_w4a16_gemmzai/glm-5.1opencode45m20260428_175639_opencode_zai_glm-5.1_07_w4a16_gemm
0.16703_paged_attentiondeepseek/deepseek-v4-flashopencode28m20260427_082406_opencode_deepseek_deepseek-v4-flash_03_paged_attention
0.16106_sonic_moe_swiglukimi-k2.6kimi45m20260428_125718_kimi_kimi-k2.6_06_sonic_moe_swiglu
0.15907_w4a16_gemmgpt-5.5 [xhigh]codex14m20260428_015550_codex_gpt-5.5_07_w4a16_gemm
0.13804_kahan_softmaxdeepseek/deepseek-v4-flashopencode16m20260427_094004_opencode_deepseek_deepseek-v4-flash_04_kahan_softmax
0.13707_w4a16_gemmor/xiaomi/mimo-v2.5-proopencode35m20260428_222337_opencode_openrouter-pinned_xiaomi_mimo-v2.5-pro_07_w4a16_gemm
0.13407_w4a16_gemmdeepseek/deepseek-v4-flashopencode18m20260427_140540_opencode_deepseek_deepseek-v4-flash_07_w4a16_gemm
0.12507_w4a16_gemmor/qwen/qwen3.6-plusopencode24m20260429_085704_opencode_openrouter-pinned_qwen_qwen3.6-plus_07_w4a16_gemm
0.12507_w4a16_gemmdeepseek/deepseek-v4-proopencode45m20260427_142403_opencode_deepseek_deepseek-v4-pro_07_w4a16_gemm
0.12504_kahan_softmaxzai/glm-5.1opencode33m20260428_162234_opencode_zai_glm-5.1_04_kahan_softmax
0.12104_kahan_softmaxor/xiaomi/mimo-v2.5-proopencode42m20260428_201818_opencode_openrouter-pinned_xiaomi_mimo-v2.5-pro_04_kahan_softmax
0.11804_kahan_softmaxkimi-k2.6kimi26m20260428_114524_kimi_kimi-k2.6_04_kahan_softmax
0.11007_w4a16_gemmor/qwen/qwen3.6-max-previewopencode45m20260429_052810_opencode_openrouter-pinned_qwen_qwen3.6-max-preview_07_w4a16_gemm
0.10806_sonic_moe_swigludeepseek/deepseek-v4-proopencode45m20260427_122352_opencode_deepseek_deepseek-v4-pro_06_sonic_moe_swiglu
0.10104_kahan_softmaxdeepseek/deepseek-v4-proopencode31m20260427_095552_opencode_deepseek_deepseek-v4-pro_04_kahan_softmax
0.08306_sonic_moe_swigludeepseek/deepseek-v4-flashopencode45m20260427_113525_opencode_deepseek_deepseek-v4-flash_06_sonic_moe_swiglu
0.08206_sonic_moe_swigluor/qwen/qwen3.6-27bopencode18m20260427_202932_opencode_openrouter-pinned_qwen_qwen3.6-27b_06_sonic_moe_swiglu
0.07704_kahan_softmaxor/qwen/qwen3.6-max-previewopencode35m20260429_002904_opencode_openrouter-pinned_qwen_qwen3.6-max-preview_04_kahan_softmax
0.07606_sonic_moe_swigluor/minimax/minimax-m2.7opencode45m20260427_131223_opencode_openrouter-pinned_minimax_minimax-m2.7_06_sonic_moe_swiglu
0.04205_topk_bitonicgpt-5.5 [xhigh]codex25m20260428_001416_codex_gpt-5.5_05_topk_bitonic
0.04006_sonic_moe_swigluor/qwen/qwen3.6-plusopencode45m20260429_080716_opencode_openrouter-pinned_qwen_qwen3.6-plus_06_sonic_moe_swiglu
0.03404_kahan_softmaxor/minimax/minimax-m2.7opencode17m20260427_102645_opencode_openrouter-pinned_minimax_minimax-m2.7_04_kahan_softmax
0.03202_kda_cutlassgpt-5.5 [xhigh]codex26m20260427_232335_codex_gpt-5.5_02_kda_cutlass
0.03007_w4a16_gemmor/minimax/minimax-m2.7opencode21m20260427_150902_opencode_openrouter-pinned_minimax_minimax-m2.7_07_w4a16_gemm
0.02703_paged_attentiondeepseek/deepseek-v4-proopencode32m20260427_085225_opencode_deepseek_deepseek-v4-pro_03_paged_attention
0.02202_kda_cutlasskimi-k2.6kimi45m20260428_095829_kimi_kimi-k2.6_02_kda_cutlass
0.02203_paged_attentionor/qwen/qwen3.6-plusopencode45m20260429_071042_opencode_openrouter-pinned_qwen_qwen3.6-plus_03_paged_attention
0.02005_topk_bitonicclaude-opus-4-7 [max]claude45m20260428_070610_claude_claude-opus-4-7_05_topk_bitonic
0.01705_topk_bitonicor/xiaomi/mimo-v2.5-proopencode35m20260428_210011_opencode_openrouter-pinned_xiaomi_mimo-v2.5-pro_05_topk_bitonic
0.01405_topk_bitonickimi-k2.6kimi45m20260428_121144_kimi_kimi-k2.6_05_topk_bitonic
0.01105_topk_bitonicdeepseek/deepseek-v4-proopencode28m20260427_105224_opencode_deepseek_deepseek-v4-pro_05_topk_bitonic
0.01102_kda_cutlassor/qwen/qwen3.6-max-previewopencode45m20260428_233804_opencode_openrouter-pinned_qwen_qwen3.6-max-preview_02_kda_cutlass
0.00902_kda_cutlassdeepseek/deepseek-v4-flashopencode45m20260427_060330_opencode_deepseek_deepseek-v4-flash_02_kda_cutlass
0.00502_kda_cutlasszai/glm-5.1opencode45m20260428_151712_opencode_zai_glm-5.1_02_kda_cutlass
0.00406_sonic_moe_swigluor/qwen/qwen3.6-max-previewopencode45m20260429_043758_opencode_openrouter-pinned_qwen_qwen3.6-max-preview_06_sonic_moe_swiglu
PASS02_kda_cutlassclaude-opus-4-7 [max]claude15m20260428_045050_claude_claude-opus-4-7_02_kda_cutlass
FAIL06_sonic_moe_swigluclaude-opus-4-7 [max]claude45m20260428_075117_claude_claude-opus-4-7_06_sonic_moe_swiglu
FAIL01_fp8_gemmkimi-k2.6kimi34m20260428_092426_kimi_kimi-k2.6_01_fp8_gemm
FAIL02_kda_cutlassor/xiaomi/mimo-v2.5-proopencode45m20260428_192328_opencode_openrouter-pinned_xiaomi_mimo-v2.5-pro_02_kda_cutlass
FAIL05_topk_bitonicor/qwen/qwen3.6-max-previewopencode45m20260429_035235_opencode_openrouter-pinned_qwen_qwen3.6-max-preview_05_topk_bitonic
FAIL05_topk_bitonicdeepseek/deepseek-v4-flashopencode9m20260427_104347_opencode_deepseek_deepseek-v4-flash_05_topk_bitonic
FAIL01_fp8_gemmdeepseek/deepseek-v4-flashopencode31m20260427_044258_opencode_deepseek_deepseek-v4-flash_01_fp8_gemm
FAIL01_fp8_gemmdeepseek/deepseek-v4-proopencode45m20260427_051426_opencode_deepseek_deepseek-v4-pro_01_fp8_gemm
FAIL02_kda_cutlassdeepseek/deepseek-v4-proopencode45m20260427_065343_opencode_deepseek_deepseek-v4-pro_02_kda_cutlass
FAIL05_topk_bitonicor/qwen/qwen3.6-plusopencode7m20260429_080043_opencode_openrouter-pinned_qwen_qwen3.6-plus_05_topk_bitonic
FAIL01_fp8_gemmzai/glm-5.1opencode45m20260428_143210_opencode_zai_glm-5.1_01_fp8_gemm
FAIL03_paged_attentionor/minimax/minimax-m2.7opencode8m20260427_092504_opencode_openrouter-pinned_minimax_minimax-m2.7_03_paged_attention
FAIL05_topk_bitonicor/minimax/minimax-m2.7opencode4m20260427_112034_opencode_openrouter-pinned_minimax_minimax-m2.7_05_topk_bitonic
FAIL02_kda_cutlassor/qwen/qwen3.6-27bopencode45m20260427_180248_opencode_openrouter-pinned_qwen_qwen3.6-27b_02_kda_cutlass
FAIL03_paged_attentionor/qwen/qwen3.6-27bopencode45m20260427_184750_opencode_openrouter-pinned_qwen_qwen3.6-27b_03_paged_attention
FAIL05_topk_bitonicor/qwen/qwen3.6-27bopencode45m20260427_194429_opencode_openrouter-pinned_qwen_qwen3.6-27b_05_topk_bitonic
ERR03_paged_attentionor/xiaomi/mimo-v2.5-proopencode10m20260428_200832_opencode_openrouter-pinned_xiaomi_mimo-v2.5-pro_03_paged_attention
ERR03_paged_attentionor/qwen/qwen3.6-max-previewopencode1m20260429_002816_opencode_openrouter-pinned_qwen_qwen3.6-max-preview_03_paged_attention
ERR02_kda_cutlassor/qwen/qwen3.6-plusopencode12m20260429_065845_opencode_openrouter-pinned_qwen_qwen3.6-plus_02_kda_cutlass
ERR04_kahan_softmaxor/qwen/qwen3.6-plusopencode5m20260429_075600_opencode_openrouter-pinned_qwen_qwen3.6-plus_04_kahan_softmax
ERR03_paged_attentionzai/glm-5.1opencode15m20260428_160715_opencode_zai_glm-5.1_03_paged_attention
ERR05_topk_bitoniczai/glm-5.1opencode13m20260428_165523_opencode_zai_glm-5.1_05_topk_bitonic
ERR01_fp8_gemmor/minimax/minimax-m2.7opencode3m20260427_055936_opencode_openrouter-pinned_minimax_minimax-m2.7_01_fp8_gemm
ERR02_kda_cutlassor/minimax/minimax-m2.7opencode45m20260427_073847_opencode_openrouter-pinned_minimax_minimax-m2.7_02_kda_cutlass
ERR01_fp8_gemmor/qwen/qwen3.6-27bopencode7m20260427_175551_opencode_openrouter-pinned_qwen_qwen3.6-27b_01_fp8_gemm
ERR04_kahan_softmaxor/qwen/qwen3.6-27bopencode12m20260427_193251_opencode_openrouter-pinned_qwen_qwen3.6-27b_04_kahan_softmax
ERR07_w4a16_gemmor/qwen/qwen3.6-27bopencode0m20260427_205128_opencode_openrouter-pinned_qwen_qwen3.6-27b_07_w4a16_gemm
ERR01_fp8_gemmor/qwen/qwen3.6-35b-a3bopencode0m20260429_092114_opencode_openrouter-pinned_qwen_qwen3.6-35b-a3b_01_fp8_gemm
ERR02_kda_cutlassor/qwen/qwen3.6-35b-a3bopencode0m20260429_092115_opencode_openrouter-pinned_qwen_qwen3.6-35b-a3b_02_kda_cutlass
ERR03_paged_attentionor/qwen/qwen3.6-35b-a3bopencode0m20260429_092116_opencode_openrouter-pinned_qwen_qwen3.6-35b-a3b_03_paged_attention
ERR04_kahan_softmaxor/qwen/qwen3.6-35b-a3bopencode0m20260429_092118_opencode_openrouter-pinned_qwen_qwen3.6-35b-a3b_04_kahan_softmax
ERR05_topk_bitonicor/qwen/qwen3.6-35b-a3bopencode0m20260429_092119_opencode_openrouter-pinned_qwen_qwen3.6-35b-a3b_05_topk_bitonic
ERR06_sonic_moe_swigluor/qwen/qwen3.6-35b-a3bopencode0m20260429_092120_opencode_openrouter-pinned_qwen_qwen3.6-35b-a3b_06_sonic_moe_swiglu
ERR07_w4a16_gemmor/qwen/qwen3.6-35b-a3bopencode0m20260429_092121_opencode_openrouter-pinned_qwen_qwen3.6-35b-a3b_07_w4a16_gemm