kernelbench.com

all runs

53 runs · 50 scored · 0 no perf · 3 fail · 0 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
0.67703_paged_attentionglm-5.2Claude Code234m20260613_055815_zai-claude_glm-5.2_03_paged_attention
0.67103_paged_attentionclaude-opus-4-8Claude Code112m20260613_042305_claude_claude-opus-4-8_03_paged_attention
0.63003_paged_attentionclaude-fable-5 [max]Claude Code61m20260611_050147_claude_claude-fable-5_03_paged_attention
0.55603_paged_attentiongpt-5.5 [xhigh]codex27m20260613_044538_codex_gpt-5.5_03_paged_attention
0.51303_paged_attentionMiniMax-M3Claude Code355m20260613_081306_minimax-claude_MiniMax-M3_03_paged_attention
0.40601_fp8_gemmglm-5.2Claude Code96m20260614_145529_zai-claude_glm-5.2_01_fp8_gemm
0.39303_paged_attentiondeepseek-v4-prodeepseek-claude25m20260615_125721_deepseek-claude_deepseek-v4-pro_03_paged_attention
0.38601_fp8_gemmclaude-opus-4-8Claude Code50m20260614_144216_claude_claude-opus-4-8_01_fp8_gemm
0.37701_fp8_gemmcomposer-2.5-fastcursor18m20260615_114532_cursor_composer-2.5-fast_01_fp8_gemm
0.36601_fp8_gemmMiniMax-M3Claude Code169m20260614_153222_minimax-claude_MiniMax-M3_01_fp8_gemm
0.36401_fp8_gemmgpt-5.5 [xhigh]codex13m20260614_144224_codex_gpt-5.5_01_fp8_gemm
0.34807_w4a16_gemmclaude-fable-5 [max]Claude Code48m20260610_224447_claude_claude-fable-5_07_w4a16_gemm
0.34701_fp8_gemmkimi-k2.7-codekimi-claude69m20260614_171808_kimi-claude_kimi-k2.7-code_01_fp8_gemm
0.34001_fp8_gemmdeepseek-v4-prodeepseek-claude36m20260614_163128_deepseek-claude_deepseek-v4-pro_01_fp8_gemm
0.32107_w4a16_gemmglm-5.2Claude Code280m20260613_061443_zai-claude_glm-5.2_07_w4a16_gemm
0.25603_paged_attentioncomposer-2.5-fastcursor4m20260615_120402_cursor_composer-2.5-fast_03_paged_attention
0.24403_paged_attentiongemini-3.5-flashGemini CLI48m20260613_105439_gemini_gemini-3.5-flash_03_paged_attention
0.24103_paged_attentionkimi-k2.7-codekimi-claude41m20260613_163906_kimi-claude_kimi-k2.7-code_03_paged_attention
0.23507_w4a16_gemmclaude-opus-4-8Claude Code218m20260613_115410_claude_claude-opus-4-8_07_w4a16_gemm
0.20307_w4a16_gemmgpt-5.5 [xhigh]codex20m20260613_053815_codex_gpt-5.5_07_w4a16_gemm
0.19601_fp8_gemmgemini-3.5-flashGemini CLI11m20260614_170713_gemini_gemini-3.5-flash_01_fp8_gemm
0.17207_w4a16_gemmgemini-3.5-flashGemini CLI101m20260613_114239_gemini_gemini-3.5-flash_07_w4a16_gemm
0.15407_w4a16_gemmdeepseek-v4-prodeepseek-claude53m20260615_143020_deepseek-claude_deepseek-v4-pro_07_w4a16_gemm
0.15307_w4a16_gemmkimi-k2.7-codekimi-claude95m20260613_181541_kimi-claude_kimi-k2.7-code_07_w4a16_gemm
0.15107_w4a16_gemmcomposer-2.5-fastcursor24m20260615_123336_cursor_composer-2.5-fast_07_w4a16_gemm
0.14507_w4a16_gemmMiniMax-M3Claude Code280m20260613_095207_minimax-claude_MiniMax-M3_07_w4a16_gemm
0.10806_sonic_moe_swigluclaude-fable-5 [max]Claude Code55m20260611_050205_claude_claude-fable-5_06_sonic_moe_swiglu
0.10306_sonic_moe_swiglukimi-k2.7-codekimi-claude166m20260613_171946_kimi-claude_kimi-k2.7-code_06_sonic_moe_swiglu
0.10106_sonic_moe_swiglucomposer-2.5-fastcursor39m20260615_121705_cursor_composer-2.5-fast_06_sonic_moe_swiglu
0.09906_sonic_moe_swiglugpt-5.5 [xhigh]codex27m20260613_051216_codex_gpt-5.5_06_sonic_moe_swiglu
0.09806_sonic_moe_swigluglm-5.2Claude Code241m20260613_061433_zai-claude_glm-5.2_06_sonic_moe_swiglu
0.09206_sonic_moe_swigluMiniMax-M3Claude Code364m20260613_094226_minimax-claude_MiniMax-M3_06_sonic_moe_swiglu
0.08706_sonic_moe_swiglugemini-3.5-flashGemini CLI106m20260613_114210_gemini_gemini-3.5-flash_06_sonic_moe_swiglu
0.08606_sonic_moe_swigluclaude-opus-4-8Claude Code105m20260613_042321_claude_claude-opus-4-8_06_sonic_moe_swiglu
0.05502_kda_cutlassclaude-opus-4-8Claude Code114m20260613_042257_claude_claude-opus-4-8_02_kda_cutlass
0.05306_sonic_moe_swigludeepseek-v4-prodeepseek-claude97m20260615_140149_deepseek-claude_deepseek-v4-pro_06_sonic_moe_swiglu
0.04905_topk_bitonicclaude-fable-5 [max]Claude Code48m20260610_174700_claude_claude-fable-5_05_topk_bitonic
0.04605_topk_bitonicgpt-5.5 [xhigh]codex35m20260613_050335_codex_gpt-5.5_05_topk_bitonic
0.03702_kda_cutlassgpt-5.5 [xhigh]codex77m20260613_042345_codex_gpt-5.5_02_kda_cutlass
0.03602_kda_cutlassclaude-fable-5 [max]Claude Code32m20260610_220314_claude_claude-fable-5_02_kda_cutlass
0.03405_topk_bitonicglm-5.2Claude Code236m20260613_060814_zai-claude_glm-5.2_05_topk_bitonic
0.03402_kda_cutlasscomposer-2.5-fastcursor31m20260615_114540_cursor_composer-2.5-fast_02_kda_cutlass
0.03405_topk_bitonicclaude-opus-4-8Claude Code137m20260613_042313_claude_claude-opus-4-8_05_topk_bitonic
0.03202_kda_cutlassglm-5.2Claude Code241m20260613_054111_zai-claude_glm-5.2_02_kda_cutlass
0.02905_topk_bitonicgemini-3.5-flashGemini CLI85m20260613_112732_gemini_gemini-3.5-flash_05_topk_bitonic
0.01702_kda_cutlasskimi-k2.7-codekimi-claude128m20260613_163858_kimi-claude_kimi-k2.7-code_02_kda_cutlass
0.01405_topk_bitonicdeepseek-v4-prodeepseek-claude68m20260615_132230_deepseek-claude_deepseek-v4-pro_05_topk_bitonic
0.00602_kda_cutlassgemini-3.5-flashGemini CLI87m20260613_101539_gemini_gemini-3.5-flash_02_kda_cutlass
0.00605_topk_bitonicMiniMax-M3Claude Code362m20260613_081745_minimax-claude_MiniMax-M3_05_topk_bitonic
0.00405_topk_bitoniccomposer-2.5-fastcursor26m20260615_120752_cursor_composer-2.5-fast_05_topk_bitonic
FAIL02_kda_cutlassdeepseek-v4-prodeepseek-claude66m20260615_125550_deepseek-claude_deepseek-v4-pro_02_kda_cutlass
FAIL05_topk_bitonickimi-k2.7-codekimi-claude183m20260613_163914_kimi-claude_kimi-k2.7-code_05_topk_bitonic
FAIL02_kda_cutlassMiniMax-M3Claude Code364m20260613_064019_minimax-claude_MiniMax-M3_02_kda_cutlass