Leaderboard

region iad1all timeupdated 9/14/202614 model configuration(s) rankedRe-weighting of old benchmarks · policy v5N=96,987 runs

Best all-rounder

83.7%

GPT-5.6 TerraMax

OpenAI

policy v5 · 4/4 categories · 7/7 active panels · 1,962 scored units

Best intelligence

95.8%

Google: Gemini 3.8 FlashHigh

OpenRouter

policy v5 · 1/1 active panels · 198 scored units · newest panel Jul 9, 2026

Best coding

84.4%

GPT-5.6 TerraMax

OpenAI

policy v5 · 2/2 active panels · 393 scored units · newest panel Aug 3, 2026

Best instruction

98.4%

GPT-5.6 SolMax

OpenAI

policy v5 · 2/2 active panels · 534 scored units · newest panel Jul 29, 2026

Best agentic

69.5%

Z.ai: GLM 5.3 FlashMax

OpenRouter

policy v5 · 2/2 active panels · 80 scored units · newest panel Sep 5, 2026

Model											Actions
GPT-5.6 TerraMax OpenAI Current leaderboard requirements Every active score-policy benchmark must meet its scored-run and score-unit coverage requirement in this window. gpqa-diamondincluded Passed:100% / 95.0% score-unit coverage at 3 scored runs per score unit ifevalincluded Passed:100% / 95.0% score-unit coverage at 3 scored runs per score unit logicif-frontierincluded Passed:100% / 100% score-unit coverage at 1 scored runs per score unit humaneval-plusincluded Passed:95.1% / 95.0% score-unit coverage at 3 scored runs per score unit livecodebench-plus-hardincluded Passed:100% / 100% score-unit coverage at 1 scored runs per score unit bfcl-liteincluded Passed:100% / 95.0% score-unit coverage at 3 scored runs per score unit thinkingbox-benchincluded Passed:100% / 100% score-unit coverage at 1 scored runs per score unit	83.7%^† Re-weighting of old benchmarks · policy v5 Activated Sep 6, 2026. Every category is required. Open scoring policy intelligence32.1%^† 91.8%^† score × 35.0% weight = 32.1% 1/1 benchmarks · 385 scored units instruction9.5%^† 94.8%^† score × 10.0% weight = 9.5% 2/2 benchmarks · 1044 scored units coding23.2%^† 84.4%^† score × 27.5% weight = 23.2% 2/2 benchmarks · 393 scored units agentic18.9%^† 68.6%^† score × 27.5% weight = 18.9% 2/2 benchmarks · 140 scored units	91.8%^† intelligence category 91.8%^† Weighted sum of the benchmark contributions below. 91.8%^† category score × 35.0% all-rounder weight = 32.1% gpqa-diamond91.8%^† 91.8%^† score × 100% weight = 91.8% Mean verifier score · 2 full-panel executions averaged · 385/396 scored units · 100% score-unit run coverage Panel v1 · 95% interval 88.8%–95.4% · generated Jul 9, 2026 Open latest contributing execution	84.4%^† coding category 84.4%^† Weighted sum of the benchmark contributions below. 84.4%^† category score × 27.5% all-rounder weight = 23.2% humaneval-plus37.5%^† 93.7%^† score × 40.0% weight = 37.5% Mean verifier score · 2 full-panel executions averaged · 324/328 scored units · 95.1% score-unit run coverage Panel v1 · 95% interval 93.8%–99.1% · generated Jul 9, 2026 Open latest contributing execution livecodebench-plus-hard47.0% 78.3% score × 60.0% weight = 47.0% Mean verifier score · 1 full-panel execution averaged · 69/69 scored units · 100% score-unit run coverage Panel v1 · 95% interval 68.1%–87.0% · generated Aug 3, 2026 Open latest contributing execution	94.8%^† instruction category 94.8%^† Weighted sum of the benchmark contributions below. 94.8%^† category score × 10.0% all-rounder weight = 9.5% ifeval27.8%^† 92.5%^† score × 30.0% weight = 27.8% Mean verifier score · 2 full-panel executions averaged · 1020/1020 scored units · 100% score-unit run coverage Panel v1 · 95% interval 92.5%–96.1% · generated Jul 9, 2026 Open latest contributing execution logicif-frontier67.1% 95.8% score × 70.0% weight = 67.1% Strict function pass rate · 1 full-panel execution averaged · 24/24 scored units · 100% score-unit run coverage Panel v3 · 95% interval 87.5%–100% · generated Jul 29, 2026 Open latest contributing execution	68.6%^† agentic category 68.6%^† Weighted sum of the benchmark contributions below. 68.6%^† category score × 27.5% all-rounder weight = 18.9% bfcl-lite23.1%^† 76.9%^† score × 30.0% weight = 23.1% Mean verifier score · 2 full-panel executions averaged · 120/120 scored units · 100% score-unit run coverage Panel v1 · 95% interval 76.9%–76.9% · generated Jul 9, 2026 Open latest contributing execution thinkingbox-bench45.5% 65.0% score × 70.0% weight = 45.5% Mean task success · 1 full-panel execution averaged · 20/20 scored units · 100% score-unit run coverage Panel v1 · 95% interval 40.0%–85.0% · generated Sep 5, 2026 Open latest contributing execution	3.73s	6.09s	~342	183.9s	0.0%	compare
Google: Gemini 3.8 FlashHigh OpenRouter Current leaderboard requirements Every active score-policy benchmark must meet its scored-run and score-unit coverage requirement in this window. gpqa-diamondincluded Passed:100% / 95.0% score-unit coverage at 3 scored runs per score unit ifevalincluded Passed:100% / 95.0% score-unit coverage at 3 scored runs per score unit logicif-frontierincluded Passed:100% / 100% score-unit coverage at 1 scored runs per score unit humaneval-plusincluded Passed:100% / 95.0% score-unit coverage at 3 scored runs per score unit livecodebench-plus-hardincluded Passed:100% / 100% score-unit coverage at 1 scored runs per score unit bfcl-liteincluded Passed:100% / 95.0% score-unit coverage at 3 scored runs per score unit thinkingbox-benchincluded Passed:100% / 100% score-unit coverage at 1 scored runs per score unit	82.9%^† Re-weighting of old benchmarks · policy v5 Activated Sep 6, 2026. Every category is required. Open scoring policy intelligence33.5%^† 95.8%^† score × 35.0% weight = 33.5% 1/1 benchmarks · 198 scored units instruction9.3%^† 93.1%^† score × 10.0% weight = 9.3% 2/2 benchmarks · 534 scored units coding22.7%^† 82.5%^† score × 27.5% weight = 22.7% 2/2 benchmarks · 233 scored units agentic17.4%^† 63.2%^† score × 27.5% weight = 17.4% 2/2 benchmarks · 80 scored units	95.8%^† intelligence category 95.8%^† Weighted sum of the benchmark contributions below. 95.8%^† category score × 35.0% all-rounder weight = 33.5% gpqa-diamond95.8%^† 95.8%^† score × 100% weight = 95.8% Mean verifier score · 1 full-panel execution averaged · 198/198 scored units · 100% score-unit run coverage Panel v1 · 95% interval 92.9%–98.1% · generated Jul 9, 2026 Open latest contributing execution	82.5%^† coding category 82.5%^† Weighted sum of the benchmark contributions below. 82.5%^† category score × 27.5% all-rounder weight = 22.7% humaneval-plus37.3%^† 93.3%^† score × 40.0% weight = 37.3% Mean verifier score · 1 full-panel execution averaged · 164/164 scored units · 100% score-unit run coverage Panel v1 · 95% interval 89.6%–96.7% · generated Jul 9, 2026 Open latest contributing execution livecodebench-plus-hard45.2% 75.4% score × 60.0% weight = 45.2% Mean verifier score · 1 full-panel execution averaged · 69/69 scored units · 100% score-unit run coverage Panel v1 · 95% interval 65.2%–85.5% · generated Aug 3, 2026 Open latest contributing execution	93.1%^† instruction category 93.1%^† Weighted sum of the benchmark contributions below. 93.1%^† category score × 10.0% all-rounder weight = 9.3% ifeval29.0%^† 96.5%^† score × 30.0% weight = 29.0% Mean verifier score · 1 full-panel execution averaged · 510/510 scored units · 100% score-unit run coverage Panel v1 · 95% interval 94.9%–97.8% · generated Jul 9, 2026 Open latest contributing execution logicif-frontier64.2% 91.7% score × 70.0% weight = 64.2% Strict function pass rate · 1 full-panel execution averaged · 24/24 scored units · 100% score-unit run coverage Panel v3 · 95% interval 79.2%–100% · generated Jul 29, 2026 Open latest contributing execution	63.2%^† agentic category 63.2%^† Weighted sum of the benchmark contributions below. 63.2%^† category score × 27.5% all-rounder weight = 17.4% bfcl-lite24.7%^† 82.2%^† score × 30.0% weight = 24.7% Mean verifier score · 1 full-panel execution averaged · 60/60 scored units · 100% score-unit run coverage Panel v1 · 95% interval 71.7%–90.6% · generated Jul 9, 2026 Open latest contributing execution thinkingbox-bench38.5% 55.0% score × 70.0% weight = 38.5% Mean task success · 1 full-panel execution averaged · 20/20 scored units · 100% score-unit run coverage Panel v1 · 95% interval 30.0%–75.0% · generated Sep 5, 2026 Open latest contributing execution	1.52s	8.11s	~221	91.0s	0.0%	compare
GPT-5.6 SolMax OpenAI Current leaderboard requirements Every active score-policy benchmark must meet its scored-run and score-unit coverage requirement in this window. gpqa-diamondincluded Passed:98.0% / 95.0% score-unit coverage at 3 scored runs per score unit ifevalincluded Passed:100% / 95.0% score-unit coverage at 3 scored runs per score unit logicif-frontierincluded Passed:100% / 100% score-unit coverage at 1 scored runs per score unit humaneval-plusincluded Passed:100% / 95.0% score-unit coverage at 3 scored runs per score unit livecodebench-plus-hardincluded Passed:100% / 100% score-unit coverage at 1 scored runs per score unit bfcl-liteincluded Passed:100% / 95.0% score-unit coverage at 3 scored runs per score unit thinkingbox-benchincluded Passed:100% / 100% score-unit coverage at 1 scored runs per score unit	81.6%^† Re-weighting of old benchmarks · policy v5 Activated Sep 6, 2026. Every category is required. Open scoring policy intelligence32.5%^† 92.8%^† score × 35.0% weight = 32.5% 1/1 benchmarks · 197 scored units instruction9.8%^† 98.4%^† score × 10.0% weight = 9.8% 2/2 benchmarks · 534 scored units coding23.1%^† 83.9%^† score × 27.5% weight = 23.1% 2/2 benchmarks · 233 scored units agentic16.2%^† 59.0%^† score × 27.5% weight = 16.2% 2/2 benchmarks · 80 scored units	92.8%^† intelligence category 92.8%^† Weighted sum of the benchmark contributions below. 92.8%^† category score × 35.0% all-rounder weight = 32.5% gpqa-diamond92.8%^† 92.8%^† score × 100% weight = 92.8% Mean verifier score · 1 full-panel execution averaged · 197/198 scored units · 98.0% score-unit run coverage Panel v1 · 95% interval 92.8%–92.8% · generated Jul 9, 2026 Open latest contributing execution	83.9%^† coding category 83.9%^† Weighted sum of the benchmark contributions below. 83.9%^† category score × 27.5% all-rounder weight = 23.1% humaneval-plus37.8%^† 94.5%^† score × 40.0% weight = 37.8% Mean verifier score · 1 full-panel execution averaged · 164/164 scored units · 100% score-unit run coverage Panel v1 · 95% interval 94.5%–94.5% · generated Jul 9, 2026 Open latest contributing execution livecodebench-plus-hard46.1% 76.8% score × 60.0% weight = 46.1% Mean verifier score · 1 full-panel execution averaged · 69/69 scored units · 100% score-unit run coverage Panel v1 · 95% interval 66.7%–87.0% · generated Aug 3, 2026 Open latest contributing execution	98.4%^† instruction category 98.4%^† Weighted sum of the benchmark contributions below. 98.4%^† category score × 10.0% all-rounder weight = 9.8% ifeval28.4%^† 94.8%^† score × 30.0% weight = 28.4% Mean verifier score · 1 full-panel execution averaged · 510/510 scored units · 100% score-unit run coverage Panel v1 · 95% interval 94.8%–94.8% · generated Jul 9, 2026 Open latest contributing execution logicif-frontier70.0% 100% score × 70.0% weight = 70.0% Strict function pass rate · 1 full-panel execution averaged · 24/24 scored units · 100% score-unit run coverage Panel v3 · 95% interval 100%–100% · generated Jul 29, 2026 Open latest contributing execution	59.0%^† agentic category 59.0%^† Weighted sum of the benchmark contributions below. 59.0%^† category score × 27.5% all-rounder weight = 16.2% bfcl-lite24.0%^† 80.0%^† score × 30.0% weight = 24.0% Mean verifier score · 1 full-panel execution averaged · 60/60 scored units · 100% score-unit run coverage Panel v1 · 95% interval 80.0%–80.0% · generated Jul 9, 2026 Open latest contributing execution thinkingbox-bench35.0% 50.0% score × 70.0% weight = 35.0% Mean task success · 1 full-panel execution averaged · 20/20 scored units · 100% score-unit run coverage Panel v1 · 95% interval 30.0%–75.0% · generated Sep 5, 2026 Open latest contributing execution	4.47s	5.22s	~289	598.5s	0.5%	compare
GPT-5.6 LunaMax OpenAI Current leaderboard requirements Every active score-policy benchmark must meet its scored-run and score-unit coverage requirement in this window. gpqa-diamondincluded Passed:96.0% / 95.0% score-unit coverage at 3 scored runs per score unit ifevalincluded Passed:100% / 95.0% score-unit coverage at 3 scored runs per score unit logicif-frontierincluded Passed:100% / 100% score-unit coverage at 1 scored runs per score unit humaneval-plusincluded Passed:100% / 95.0% score-unit coverage at 3 scored runs per score unit livecodebench-plus-hardincluded Passed:100% / 100% score-unit coverage at 1 scored runs per score unit bfcl-liteincluded Passed:100% / 95.0% score-unit coverage at 3 scored runs per score unit thinkingbox-benchincluded Passed:100% / 100% score-unit coverage at 1 scored runs per score unit	77.7%^† Re-weighting of old benchmarks · policy v5 Activated Sep 6, 2026. Every category is required. Open scoring policy intelligence30.1%^† 86.0%^† score × 35.0% weight = 30.1% 1/1 benchmarks · 196 scored units instruction7.7%^† 77.0%^† score × 10.0% weight = 7.7% 2/2 benchmarks · 534 scored units coding22.6%^† 82.2%^† score × 27.5% weight = 22.6% 2/2 benchmarks · 233 scored units agentic17.2%^† 62.7%^† score × 27.5% weight = 17.2% 2/2 benchmarks · 80 scored units	86.0%^† intelligence category 86.0%^† Weighted sum of the benchmark contributions below. 86.0%^† category score × 35.0% all-rounder weight = 30.1% gpqa-diamond86.0%^† 86.0%^† score × 100% weight = 86.0% Mean verifier score · 1 full-panel execution averaged · 196/198 scored units · 96.0% score-unit run coverage Panel v1 · 95% interval 86.0%–86.0% · generated Jul 9, 2026 Open latest contributing execution	82.2%^† coding category 82.2%^† Weighted sum of the benchmark contributions below. 82.2%^† category score × 27.5% all-rounder weight = 22.6% humaneval-plus37.0%^† 92.6%^† score × 40.0% weight = 37.0% Mean verifier score · 1 full-panel execution averaged · 164/164 scored units · 100% score-unit run coverage Panel v1 · 95% interval 92.6%–92.6% · generated Jul 9, 2026 Open latest contributing execution livecodebench-plus-hard45.2% 75.4% score × 60.0% weight = 45.2% Mean verifier score · 1 full-panel execution averaged · 69/69 scored units · 100% score-unit run coverage Panel v1 · 95% interval 65.2%–84.1% · generated Aug 3, 2026 Open latest contributing execution	77.0%^† instruction category 77.0%^† Weighted sum of the benchmark contributions below. 77.0%^† category score × 10.0% all-rounder weight = 7.7% ifeval27.4%^† 91.3%^† score × 30.0% weight = 27.4% Mean verifier score · 1 full-panel execution averaged · 510/510 scored units · 100% score-unit run coverage Panel v1 · 95% interval 91.3%–91.3% · generated Jul 9, 2026 Open latest contributing execution logicif-frontier49.6% 70.8% score × 70.0% weight = 49.6% Strict function pass rate · 1 full-panel execution averaged · 24/24 scored units · 100% score-unit run coverage Panel v3 · 95% interval 54.2%–87.5% · generated Jul 29, 2026 Open latest contributing execution	62.7%^† agentic category 62.7%^† Weighted sum of the benchmark contributions below. 62.7%^† category score × 27.5% all-rounder weight = 17.2% bfcl-lite24.2%^† 80.7%^† score × 30.0% weight = 24.2% Mean verifier score · 1 full-panel execution averaged · 60/60 scored units · 100% score-unit run coverage Panel v1 · 95% interval 80.7%–80.7% · generated Jul 9, 2026 Open latest contributing execution thinkingbox-bench38.5% 55.0% score × 70.0% weight = 38.5% Mean task success · 1 full-panel execution averaged · 20/20 scored units · 100% score-unit run coverage Panel v1 · 95% interval 30.0%–75.0% · generated Sep 5, 2026 Open latest contributing execution	3.12s	4.67s	~649	437.2s	0.1%	compare
Claude Opus 5Extra high Anthropic Current leaderboard requirements Every active score-policy benchmark must meet its scored-run and score-unit coverage requirement in this window. gpqa-diamondincluded Passed:100% / 95.0% score-unit coverage at 3 scored runs per score unit ifevalincluded Passed:100% / 95.0% score-unit coverage at 3 scored runs per score unit logicif-frontierincluded Passed:100% / 100% score-unit coverage at 1 scored runs per score unit humaneval-plusincluded Passed:100% / 95.0% score-unit coverage at 3 scored runs per score unit livecodebench-plus-hardincluded Passed:100% / 100% score-unit coverage at 1 scored runs per score unit bfcl-liteincluded Passed:100% / 95.0% score-unit coverage at 3 scored runs per score unit thinkingbox-benchincluded Passed:100% / 100% score-unit coverage at 1 scored runs per score unit	73.2%^† Re-weighting of old benchmarks · policy v5 Activated Sep 6, 2026. Every category is required. Open scoring policy intelligence32.5%^† 92.8%^† score × 35.0% weight = 32.5% 1/1 benchmarks · 198 scored units instruction5.6%^† 56.1%^† score × 10.0% weight = 5.6% 2/2 benchmarks · 534 scored units coding16.9%^† 61.4%^† score × 27.5% weight = 16.9% 2/2 benchmarks · 233 scored units agentic18.2%^† 66.2%^† score × 27.5% weight = 18.2% 2/2 benchmarks · 80 scored units	92.8%^† intelligence category 92.8%^† Weighted sum of the benchmark contributions below. 92.8%^† category score × 35.0% all-rounder weight = 32.5% gpqa-diamond92.8%^† 92.8%^† score × 100% weight = 92.8% Mean verifier score · 1 full-panel execution averaged · 198/198 scored units · 100% score-unit run coverage Panel v1 · 95% interval 89.4%–95.8% · generated Jul 9, 2026 Open latest contributing execution	61.4%^† coding category 61.4%^† Weighted sum of the benchmark contributions below. 61.4%^† category score × 27.5% all-rounder weight = 16.9% humaneval-plus37.9%^† 94.7%^† score × 40.0% weight = 37.9% Mean verifier score · 1 full-panel execution averaged · 164/164 scored units · 100% score-unit run coverage Panel v1 · 95% interval 91.1%–97.8% · generated Jul 9, 2026 Open latest contributing execution livecodebench-plus-hard23.5% 39.1% score × 60.0% weight = 23.5% Mean verifier score · 1 full-panel execution averaged · 69/69 scored units · 100% score-unit run coverage Panel v1 · 95% interval 29.0%–50.7% · generated Aug 3, 2026 Open latest contributing execution	56.1%^† instruction category 56.1%^† Weighted sum of the benchmark contributions below. 56.1%^† category score × 10.0% all-rounder weight = 5.6% ifeval27.0%^† 89.9%^† score × 30.0% weight = 27.0% Mean verifier score · 1 full-panel execution averaged · 510/510 scored units · 100% score-unit run coverage Panel v1 · 95% interval 87.7%–92.1% · generated Jul 9, 2026 Open latest contributing execution logicif-frontier29.2% 41.7% score × 70.0% weight = 29.2% Strict function pass rate · 1 full-panel execution averaged · 24/24 scored units · 100% score-unit run coverage Panel v3 · 95% interval 25.0%–62.5% · generated Jul 29, 2026 Open latest contributing execution	66.2%^† agentic category 66.2%^† Weighted sum of the benchmark contributions below. 66.2%^† category score × 27.5% all-rounder weight = 18.2% bfcl-lite24.2%^† 80.6%^† score × 30.0% weight = 24.2% Mean verifier score · 1 full-panel execution averaged · 60/60 scored units · 100% score-unit run coverage Panel v1 · 95% interval 69.4%–90.0% · generated Jul 9, 2026 Open latest contributing execution thinkingbox-bench42.0% 60.0% score × 70.0% weight = 42.0% Mean task success · 1 full-panel execution averaged · 20/20 scored units · 100% score-unit run coverage Panel v1 · 95% interval 35.0%–80.0% · generated Sep 5, 2026 Open latest contributing execution	5.51s	10.2s	~50.5	103.9s	0.0%	compare
DeepSeek: DeepSeek V4 Flash 0731Max OpenRouter Current leaderboard requirements Every active score-policy benchmark must meet its scored-run and score-unit coverage requirement in this window. gpqa-diamondincluded Passed:100% / 95.0% score-unit coverage at 3 scored runs per score unit ifevalincluded Passed:100% / 95.0% score-unit coverage at 3 scored runs per score unit logicif-frontierincluded Passed:100% / 100% score-unit coverage at 1 scored runs per score unit humaneval-plusincluded Passed:100% / 95.0% score-unit coverage at 3 scored runs per score unit livecodebench-plus-hardincluded Passed:100% / 100% score-unit coverage at 1 scored runs per score unit bfcl-liteincluded Passed:100% / 95.0% score-unit coverage at 3 scored runs per score unit thinkingbox-benchincluded Passed:100% / 100% score-unit coverage at 1 scored runs per score unit	70.3%^† Re-weighting of old benchmarks · policy v5 Activated Sep 6, 2026. Every category is required. Open scoring policy intelligence30.3%^† 86.7%^† score × 35.0% weight = 30.3% 1/1 benchmarks · 198 scored units instruction5.7%^† 57.0%^† score × 10.0% weight = 5.7% 2/2 benchmarks · 534 scored units coding16.1%^† 58.6%^† score × 27.5% weight = 16.1% 2/2 benchmarks · 233 scored units agentic18.2%^† 66.2%^† score × 27.5% weight = 18.2% 2/2 benchmarks · 80 scored units	86.7%^† intelligence category 86.7%^† Weighted sum of the benchmark contributions below. 86.7%^† category score × 35.0% all-rounder weight = 30.3% gpqa-diamond86.7%^† 86.7%^† score × 100% weight = 86.7% Mean verifier score · 1 full-panel execution averaged · 198/198 scored units · 100% score-unit run coverage Panel v1 · 95% interval 82.8%–90.6% · generated Jul 9, 2026 Open latest contributing execution	58.6%^† coding category 58.6%^† Weighted sum of the benchmark contributions below. 58.6%^† category score × 27.5% all-rounder weight = 16.1% humaneval-plus36.8%^† 92.1%^† score × 40.0% weight = 36.8% Mean verifier score · 1 full-panel execution averaged · 164/164 scored units · 100% score-unit run coverage Panel v1 · 95% interval 88.8%–95.3% · generated Jul 9, 2026 Open latest contributing execution livecodebench-plus-hard21.7% 36.2% score × 60.0% weight = 21.7% Mean verifier score · 1 full-panel execution averaged · 69/69 scored units · 100% score-unit run coverage Panel v1 · 95% interval 24.6%–47.8% · generated Aug 3, 2026 Open latest contributing execution	57.0%^† instruction category 57.0%^† Weighted sum of the benchmark contributions below. 57.0%^† category score × 10.0% all-rounder weight = 5.7% ifeval27.8%^† 92.6%^† score × 30.0% weight = 27.8% Mean verifier score · 1 full-panel execution averaged · 510/510 scored units · 100% score-unit run coverage Panel v1 · 95% interval 91.0%–94.3% · generated Jul 9, 2026 Open latest contributing execution logicif-frontier29.2% 41.7% score × 70.0% weight = 29.2% Strict function pass rate · 1 full-panel execution averaged · 24/24 scored units · 100% score-unit run coverage Panel v3 · 95% interval 20.8%–58.3% · generated Jul 29, 2026 Open latest contributing execution	66.2%^† agentic category 66.2%^† Weighted sum of the benchmark contributions below. 66.2%^† category score × 27.5% all-rounder weight = 18.2% bfcl-lite24.2%^† 80.6%^† score × 30.0% weight = 24.2% Mean verifier score · 1 full-panel execution averaged · 60/60 scored units · 100% score-unit run coverage Panel v1 · 95% interval 68.9%–90.0% · generated Jul 9, 2026 Open latest contributing execution thinkingbox-bench42.0% 60.0% score × 70.0% weight = 42.0% Mean task success · 1 full-panel execution averaged · 20/20 scored units · 100% score-unit run coverage Panel v1 · 95% interval 40.0%–80.0% · generated Sep 5, 2026 Open latest contributing execution	865ms	11.8s	~61.7	451.2s	0.0%	compare
Claude Sonnet 5Extra high Anthropic Current leaderboard requirements Every active score-policy benchmark must meet its scored-run and score-unit coverage requirement in this window. gpqa-diamondincluded Passed:100% / 95.0% score-unit coverage at 3 scored runs per score unit ifevalincluded Passed:96.7% / 95.0% score-unit coverage at 3 scored runs per score unit logicif-frontierincluded Passed:100% / 100% score-unit coverage at 1 scored runs per score unit humaneval-plusincluded Passed:100% / 95.0% score-unit coverage at 3 scored runs per score unit livecodebench-plus-hardincluded Passed:100% / 100% score-unit coverage at 1 scored runs per score unit bfcl-liteincluded Passed:100% / 95.0% score-unit coverage at 3 scored runs per score unit thinkingbox-benchincluded Passed:100% / 100% score-unit coverage at 1 scored runs per score unit	69.8%^† Re-weighting of old benchmarks · policy v5 Activated Sep 6, 2026. Every category is required. Open scoring policy intelligence31.5%^† 89.9%^† score × 35.0% weight = 31.5% 1/1 benchmarks · 363 scored units instruction7.0%^† 70.1%^† score × 10.0% weight = 7.0% 2/2 benchmarks · 1036 scored units coding15.0%^† 54.5%^† score × 27.5% weight = 15.0% 2/2 benchmarks · 397 scored units agentic16.3%^† 59.3%^† score × 27.5% weight = 16.3% 2/2 benchmarks · 140 scored units	89.9%^† intelligence category 89.9%^† Weighted sum of the benchmark contributions below. 89.9%^† category score × 35.0% all-rounder weight = 31.5% gpqa-diamond89.9%^† 89.9%^† score × 100% weight = 89.9% Mean verifier score · 2 full-panel executions averaged · 363/396 scored units · 100% score-unit run coverage Panel v1 · 95% interval 85.8%–94.1% · generated Jul 9, 2026 Open latest contributing execution	54.5%^† coding category 54.5%^† Weighted sum of the benchmark contributions below. 54.5%^† category score × 27.5% all-rounder weight = 15.0% humaneval-plus37.1%^† 92.7%^† score × 40.0% weight = 37.1% Mean verifier score · 2 full-panel executions averaged · 328/328 scored units · 100% score-unit run coverage Panel v1 · 95% interval 88.1%–95.7% · generated Jul 9, 2026 Open latest contributing execution livecodebench-plus-hard17.4% 29.0% score × 60.0% weight = 17.4% Mean verifier score · 1 full-panel execution averaged · 69/69 scored units · 100% score-unit run coverage Panel v1 · 95% interval 18.8%–40.6% · generated Aug 3, 2026 Open latest contributing execution	70.1%^† instruction category 70.1%^† Weighted sum of the benchmark contributions below. 70.1%^† category score × 10.0% all-rounder weight = 7.0% ifeval26.4%^† 87.9%^† score × 30.0% weight = 26.4% Mean verifier score · 2 full-panel executions averaged · 1012/1020 scored units · 96.7% score-unit run coverage Panel v1 · 95% interval 86.0%–91.3% · generated Jul 9, 2026 Open latest contributing execution logicif-frontier43.8% 62.5% score × 70.0% weight = 43.8% Strict function pass rate · 1 full-panel execution averaged · 24/24 scored units · 100% score-unit run coverage Panel v3 · 95% interval 41.7%–83.3% · generated Jul 29, 2026 Open latest contributing execution	59.3%^† agentic category 59.3%^† Weighted sum of the benchmark contributions below. 59.3%^† category score × 27.5% all-rounder weight = 16.3% bfcl-lite24.3%^† 80.8%^† score × 30.0% weight = 24.3% Mean verifier score · 2 full-panel executions averaged · 120/120 scored units · 100% score-unit run coverage Panel v1 · 95% interval 70.0%–90.0% · generated Jul 9, 2026 Open latest contributing execution thinkingbox-bench35.0% 50.0% score × 70.0% weight = 35.0% Mean task success · 1 full-panel execution averaged · 20/20 scored units · 100% score-unit run coverage Panel v1 · 95% interval 25.0%–70.0% · generated Sep 5, 2026 Open latest contributing execution	3.00s	3.76s	~106	137.3s	0.0%	compare
Kimi K3Max OpenRouter Current leaderboard requirements Every active score-policy benchmark must meet its scored-run and score-unit coverage requirement in this window. gpqa-diamondincluded Passed:100% / 95.0% score-unit coverage at 3 scored runs per score unit ifevalincluded Passed:100% / 95.0% score-unit coverage at 3 scored runs per score unit logicif-frontierincluded Passed:100% / 100% score-unit coverage at 1 scored runs per score unit humaneval-plusincluded Passed:100% / 95.0% score-unit coverage at 3 scored runs per score unit livecodebench-plus-hardincluded Passed:100% / 100% score-unit coverage at 1 scored runs per score unit bfcl-liteincluded Passed:100% / 95.0% score-unit coverage at 3 scored runs per score unit thinkingbox-benchincluded Passed:100% / 100% score-unit coverage at 1 scored runs per score unit	69.6%^† Re-weighting of old benchmarks · policy v5 Activated Sep 6, 2026. Every category is required. Open scoring policy intelligence31.6%^† 90.4%^† score × 35.0% weight = 31.6% 1/1 benchmarks · 198 scored units instruction7.8%^† 77.5%^† score × 10.0% weight = 7.8% 2/2 benchmarks · 534 scored units coding12.2%^† 44.2%^† score × 27.5% weight = 12.2% 2/2 benchmarks · 233 scored units agentic18.1%^† 65.7%^† score × 27.5% weight = 18.1% 2/2 benchmarks · 80 scored units	90.4%^† intelligence category 90.4%^† Weighted sum of the benchmark contributions below. 90.4%^† category score × 35.0% all-rounder weight = 31.6% gpqa-diamond90.4%^† 90.4%^† score × 100% weight = 90.4% Mean verifier score · 1 full-panel execution averaged · 198/198 scored units · 100% score-unit run coverage Panel v1 · 95% interval 87.2%–93.6% · generated Jul 9, 2026 Open latest contributing execution	44.2%^† coding category 44.2%^† Weighted sum of the benchmark contributions below. 44.2%^† category score × 27.5% all-rounder weight = 12.2% humaneval-plus38.1%^† 95.3%^† score × 40.0% weight = 38.1% Mean verifier score · 1 full-panel execution averaged · 164/164 scored units · 100% score-unit run coverage Panel v1 · 95% interval 92.5%–98.0% · generated Jul 9, 2026 Open latest contributing execution livecodebench-plus-hard6.1% 10.1% score × 60.0% weight = 6.1% Mean verifier score · 1 full-panel execution averaged · 69/69 scored units · 100% score-unit run coverage Panel v1 · 95% interval 2.9%–17.4% · generated Aug 3, 2026 Open latest contributing execution	77.5%^† instruction category 77.5%^† Weighted sum of the benchmark contributions below. 77.5%^† category score × 10.0% all-rounder weight = 7.8% ifeval27.9%^† 93.1%^† score × 30.0% weight = 27.9% Mean verifier score · 1 full-panel execution averaged · 510/510 scored units · 100% score-unit run coverage Panel v1 · 95% interval 91.2%–95.0% · generated Jul 9, 2026 Open latest contributing execution logicif-frontier49.6% 70.8% score × 70.0% weight = 49.6% Strict function pass rate · 1 full-panel execution averaged · 24/24 scored units · 100% score-unit run coverage Panel v3 · 95% interval 50.0%–87.5% · generated Jul 29, 2026 Open latest contributing execution	65.7%^† agentic category 65.7%^† Weighted sum of the benchmark contributions below. 65.7%^† category score × 27.5% all-rounder weight = 18.1% bfcl-lite23.7%^† 78.9%^† score × 30.0% weight = 23.7% Mean verifier score · 1 full-panel execution averaged · 60/60 scored units · 100% score-unit run coverage Panel v1 · 95% interval 67.2%–88.3% · generated Jul 9, 2026 Open latest contributing execution thinkingbox-bench42.0% 60.0% score × 70.0% weight = 42.0% Mean task success · 1 full-panel execution averaged · 20/20 scored units · 100% score-unit run coverage Panel v1 · 95% interval 35.0%–80.0% · generated Sep 5, 2026 Open latest contributing execution	2.20s	19.6s	~40.3	271.0s	0.0%	compare
DeepSeek: DeepSeek V4 Pro 0813Max OpenRouter Current leaderboard requirements Every active score-policy benchmark must meet its scored-run and score-unit coverage requirement in this window. gpqa-diamondincluded Passed:100% / 95.0% score-unit coverage at 3 scored runs per score unit ifevalincluded Passed:100% / 95.0% score-unit coverage at 3 scored runs per score unit logicif-frontierincluded Passed:100% / 100% score-unit coverage at 1 scored runs per score unit humaneval-plusincluded Passed:100% / 95.0% score-unit coverage at 3 scored runs per score unit livecodebench-plus-hardincluded Passed:100% / 100% score-unit coverage at 1 scored runs per score unit bfcl-liteincluded Passed:100% / 95.0% score-unit coverage at 3 scored runs per score unit thinkingbox-benchincluded Passed:100% / 100% score-unit coverage at 1 scored runs per score unit	68.6%^† Re-weighting of old benchmarks · policy v5 Activated Sep 6, 2026. Every category is required. Open scoring policy intelligence31.2%^† 89.2%^† score × 35.0% weight = 31.2% 1/1 benchmarks · 198 scored units instruction3.4%^† 34.0%^† score × 10.0% weight = 3.4% 2/2 benchmarks · 534 scored units coding18.8%^† 68.5%^† score × 27.5% weight = 18.8% 2/2 benchmarks · 233 scored units agentic15.1%^† 55.0%^† score × 27.5% weight = 15.1% 2/2 benchmarks · 80 scored units	89.2%^† intelligence category 89.2%^† Weighted sum of the benchmark contributions below. 89.2%^† category score × 35.0% all-rounder weight = 31.2% gpqa-diamond89.2%^† 89.2%^† score × 100% weight = 89.2% Mean verifier score · 1 full-panel execution averaged · 198/198 scored units · 100% score-unit run coverage Panel v1 · 95% interval 85.5%–92.9% · generated Jul 9, 2026 Open latest contributing execution	68.5%^† coding category 68.5%^† Weighted sum of the benchmark contributions below. 68.5%^† category score × 27.5% all-rounder weight = 18.8% humaneval-plus37.2%^† 92.9%^† score × 40.0% weight = 37.2% Mean verifier score · 1 full-panel execution averaged · 164/164 scored units · 100% score-unit run coverage Panel v1 · 95% interval 89.6%–95.7% · generated Jul 9, 2026 Open latest contributing execution livecodebench-plus-hard31.3% 52.2% score × 60.0% weight = 31.3% Mean verifier score · 1 full-panel execution averaged · 69/69 scored units · 100% score-unit run coverage Panel v1 · 95% interval 40.6%–63.8% · generated Aug 3, 2026 Open latest contributing execution	34.0%^† instruction category 34.0%^† Weighted sum of the benchmark contributions below. 34.0%^† category score × 10.0% all-rounder weight = 3.4% ifeval28.1%^† 93.7%^† score × 30.0% weight = 28.1% Mean verifier score · 1 full-panel execution averaged · 510/510 scored units · 100% score-unit run coverage Panel v1 · 95% interval 92.1%–95.3% · generated Jul 9, 2026 Open latest contributing execution logicif-frontier5.8% 8.3% score × 70.0% weight = 5.8% Strict function pass rate · 1 full-panel execution averaged · 24/24 scored units · 100% score-unit run coverage Panel v3 · 95% interval 0.0%–20.8% · generated Jul 29, 2026 Open latest contributing execution	55.0%^† agentic category 55.0%^† Weighted sum of the benchmark contributions below. 55.0%^† category score × 27.5% all-rounder weight = 15.1% bfcl-lite23.5%^† 78.3%^† score × 30.0% weight = 23.5% Mean verifier score · 1 full-panel execution averaged · 60/60 scored units · 100% score-unit run coverage Panel v1 · 95% interval 66.7%–87.8% · generated Jul 9, 2026 Open latest contributing execution thinkingbox-bench31.5% 45.0% score × 70.0% weight = 31.5% Mean task success · 1 full-panel execution averaged · 20/20 scored units · 100% score-unit run coverage Panel v1 · 95% interval 25.0%–65.0% · generated Sep 5, 2026 Open latest contributing execution	1.14s	17.6s	~97.0	424.5s	0.0%	compare
Z.ai: GLM 5.3 FlashMax OpenRouter Current leaderboard requirements Every active score-policy benchmark must meet its scored-run and score-unit coverage requirement in this window. gpqa-diamondincluded Passed:100% / 95.0% score-unit coverage at 3 scored runs per score unit ifevalincluded Passed:100% / 95.0% score-unit coverage at 3 scored runs per score unit logicif-frontierincluded Passed:100% / 100% score-unit coverage at 1 scored runs per score unit humaneval-plusincluded Passed:98.8% / 95.0% score-unit coverage at 3 scored runs per score unit livecodebench-plus-hardincluded Passed:100% / 100% score-unit coverage at 1 scored runs per score unit bfcl-liteincluded Passed:100% / 95.0% score-unit coverage at 3 scored runs per score unit thinkingbox-benchincluded Passed:100% / 100% score-unit coverage at 1 scored runs per score unit	68.1%^† Re-weighting of old benchmarks · policy v5 Activated Sep 6, 2026. Every category is required. Open scoring policy intelligence31.3%^† 89.4%^† score × 35.0% weight = 31.3% 1/1 benchmarks · 198 scored units instruction5.9%^† 58.8%^† score × 10.0% weight = 5.9% 2/2 benchmarks · 534 scored units coding11.8%^† 43.0%^† score × 27.5% weight = 11.8% 2/2 benchmarks · 233 scored units agentic19.1%^† 69.5%^† score × 27.5% weight = 19.1% 2/2 benchmarks · 80 scored units	89.4%^† intelligence category 89.4%^† Weighted sum of the benchmark contributions below. 89.4%^† category score × 35.0% all-rounder weight = 31.3% gpqa-diamond89.4%^† 89.4%^† score × 100% weight = 89.4% Mean verifier score · 1 full-panel execution averaged · 198/198 scored units · 100% score-unit run coverage Panel v1 · 95% interval 85.7%–93.3% · generated Jul 9, 2026 Open latest contributing execution	43.0%^† coding category 43.0%^† Weighted sum of the benchmark contributions below. 43.0%^† category score × 27.5% all-rounder weight = 11.8% humaneval-plus37.8%^† 94.5%^† score × 40.0% weight = 37.8% Mean verifier score · 1 full-panel execution averaged · 164/164 scored units · 98.8% score-unit run coverage Panel v1 · 95% interval 91.5%–97.2% · generated Jul 9, 2026 Open latest contributing execution livecodebench-plus-hard5.2% 8.7% score × 60.0% weight = 5.2% Mean verifier score · 1 full-panel execution averaged · 69/69 scored units · 100% score-unit run coverage Panel v1 · 95% interval 2.9%–15.9% · generated Aug 3, 2026 Open latest contributing execution	58.8%^† instruction category 58.8%^† Weighted sum of the benchmark contributions below. 58.8%^† category score × 10.0% all-rounder weight = 5.9% ifeval26.7%^† 89.1%^† score × 30.0% weight = 26.7% Mean verifier score · 1 full-panel execution averaged · 510/510 scored units · 100% score-unit run coverage Panel v1 · 95% interval 86.9%–91.3% · generated Jul 9, 2026 Open latest contributing execution logicif-frontier32.1% 45.8% score × 70.0% weight = 32.1% Strict function pass rate · 1 full-panel execution averaged · 24/24 scored units · 100% score-unit run coverage Panel v3 · 95% interval 25.0%–66.7% · generated Jul 29, 2026 Open latest contributing execution	69.5%^† agentic category 69.5%^† Weighted sum of the benchmark contributions below. 69.5%^† category score × 27.5% all-rounder weight = 19.1% bfcl-lite24.0%^† 80.0%^† score × 30.0% weight = 24.0% Mean verifier score · 1 full-panel execution averaged · 60/60 scored units · 100% score-unit run coverage Panel v1 · 95% interval 68.3%–88.9% · generated Jul 9, 2026 Open latest contributing execution thinkingbox-bench45.5% 65.0% score × 70.0% weight = 45.5% Mean task success · 1 full-panel execution averaged · 20/20 scored units · 100% score-unit run coverage Panel v1 · 95% interval 45.0%–85.0% · generated Sep 5, 2026 Open latest contributing execution	1.32s	25.0s	~70.5	537.8s	0.6%	compare
Ling-3.0-flashAdaptive OpenRouter Current leaderboard requirements Every active score-policy benchmark must meet its scored-run and score-unit coverage requirement in this window. gpqa-diamondincluded Passed:98.0% / 95.0% score-unit coverage at 3 scored runs per score unit ifevalincluded Passed:100% / 95.0% score-unit coverage at 3 scored runs per score unit logicif-frontierincluded Passed:100% / 100% score-unit coverage at 1 scored runs per score unit humaneval-plusincluded Passed:100% / 95.0% score-unit coverage at 3 scored runs per score unit livecodebench-plus-hardincluded Passed:100% / 100% score-unit coverage at 1 scored runs per score unit bfcl-liteincluded Passed:100% / 95.0% score-unit coverage at 3 scored runs per score unit thinkingbox-benchincluded Passed:100% / 100% score-unit coverage at 1 scored runs per score unit	50.4%^† Re-weighting of old benchmarks · policy v5 Activated Sep 6, 2026. Every category is required. Open scoring policy intelligence24.5%^† 70.0%^† score × 35.0% weight = 24.5% 1/1 benchmarks · 198 scored units instruction2.8%^† 28.2%^† score × 10.0% weight = 2.8% 2/2 benchmarks · 534 scored units coding10.6%^† 38.7%^† score × 27.5% weight = 10.6% 2/2 benchmarks · 233 scored units agentic12.5%^† 45.3%^† score × 27.5% weight = 12.5% 2/2 benchmarks · 80 scored units	70.0%^† intelligence category 70.0%^† Weighted sum of the benchmark contributions below. 70.0%^† category score × 35.0% all-rounder weight = 24.5% gpqa-diamond70.0%^† 70.0%^† score × 100% weight = 70.0% Mean verifier score · 1 full-panel execution averaged · 198/198 scored units · 98.0% score-unit run coverage Panel v1 · 95% interval 64.5%–75.4% · generated Jul 9, 2026 Open latest contributing execution	38.7%^† coding category 38.7%^† Weighted sum of the benchmark contributions below. 38.7%^† category score × 27.5% all-rounder weight = 10.6% humaneval-plus37.8%^† 94.5%^† score × 40.0% weight = 37.8% Mean verifier score · 1 full-panel execution averaged · 164/164 scored units · 100% score-unit run coverage Panel v1 · 95% interval 91.7%–97.4% · generated Jul 9, 2026 Open latest contributing execution livecodebench-plus-hard0.9% 1.4% score × 60.0% weight = 0.9% Mean verifier score · 1 full-panel execution averaged · 69/69 scored units · 100% score-unit run coverage Panel v1 · 95% interval 0.0%–4.3% · generated Aug 3, 2026 Open latest contributing execution	28.2%^† instruction category 28.2%^† Weighted sum of the benchmark contributions below. 28.2%^† category score × 10.0% all-rounder weight = 2.8% ifeval28.2%^† 94.0%^† score × 30.0% weight = 28.2% Mean verifier score · 1 full-panel execution averaged · 510/510 scored units · 100% score-unit run coverage Panel v1 · 95% interval 92.3%–95.5% · generated Jul 9, 2026 Open latest contributing execution logicif-frontier0.0% 0.0% score × 70.0% weight = 0.0% Strict function pass rate · 1 full-panel execution averaged · 24/24 scored units · 100% score-unit run coverage Panel v3 · 95% interval 0.0%–0.0% · generated Jul 29, 2026 Open latest contributing execution	45.3%^† agentic category 45.3%^† Weighted sum of the benchmark contributions below. 45.3%^† category score × 27.5% all-rounder weight = 12.5% bfcl-lite24.3%^† 81.1%^† score × 30.0% weight = 24.3% Mean verifier score · 1 full-panel execution averaged · 60/60 scored units · 100% score-unit run coverage Panel v1 · 95% interval 69.4%–90.0% · generated Jul 9, 2026 Open latest contributing execution thinkingbox-bench21.0% 30.0% score × 70.0% weight = 21.0% Mean task success · 1 full-panel execution averaged · 20/20 scored units · 100% score-unit run coverage Panel v1 · 95% interval 10.0%–50.0% · generated Sep 5, 2026 Open latest contributing execution	745ms	4.81s	~442	102.7s	0.0%	compare
Claude Haiku 4.5Max Anthropic Current leaderboard requirements Every active score-policy benchmark must meet its scored-run and score-unit coverage requirement in this window. gpqa-diamondincluded Passed:100% / 95.0% score-unit coverage at 3 scored runs per score unit ifevalincluded Passed:100% / 95.0% score-unit coverage at 3 scored runs per score unit logicif-frontierincluded Passed:100% / 100% score-unit coverage at 1 scored runs per score unit humaneval-plusincluded Passed:100% / 95.0% score-unit coverage at 3 scored runs per score unit livecodebench-plus-hardincluded Passed:100% / 100% score-unit coverage at 1 scored runs per score unit bfcl-liteincluded Passed:100% / 95.0% score-unit coverage at 3 scored runs per score unit thinkingbox-benchincluded Passed:100% / 100% score-unit coverage at 1 scored runs per score unit	45.1%^† Re-weighting of old benchmarks · policy v5 Activated Sep 6, 2026. Every category is required. Open scoring policy intelligence23.8%^† 67.9%^† score × 35.0% weight = 23.8% 1/1 benchmarks · 395 scored units instruction3.5%^† 34.5%^† score × 10.0% weight = 3.5% 2/2 benchmarks · 1044 scored units coding11.1%^† 40.2%^† score × 27.5% weight = 11.1% 2/2 benchmarks · 397 scored units agentic6.8%^† 24.8%^† score × 27.5% weight = 6.8% 2/2 benchmarks · 200 scored units	67.9%^† intelligence category 67.9%^† Weighted sum of the benchmark contributions below. 67.9%^† category score × 35.0% all-rounder weight = 23.8% gpqa-diamond67.9%^† 67.9%^† score × 100% weight = 67.9% Mean verifier score · 2 full-panel executions averaged · 395/396 scored units · 100% score-unit run coverage Panel v1 · 95% interval 66.0%–78.7% · generated Jul 9, 2026 Open latest contributing execution	40.2%^† coding category 40.2%^† Weighted sum of the benchmark contributions below. 40.2%^† category score × 27.5% all-rounder weight = 11.1% humaneval-plus36.8%^† 91.9%^† score × 40.0% weight = 36.8% Mean verifier score · 2 full-panel executions averaged · 328/328 scored units · 100% score-unit run coverage Panel v1 · 95% interval 89.0%–97.0% · generated Jul 9, 2026 Open latest contributing execution livecodebench-plus-hard3.5% 5.8% score × 60.0% weight = 3.5% Mean verifier score · 1 full-panel execution averaged · 69/69 scored units · 100% score-unit run coverage Panel v1 · 95% interval 1.4%–11.6% · generated Aug 3, 2026 Open latest contributing execution	34.5%^† instruction category 34.5%^† Weighted sum of the benchmark contributions below. 34.5%^† category score × 10.0% all-rounder weight = 3.5% ifeval25.8%^† 86.0%^† score × 30.0% weight = 25.8% Mean verifier score · 2 full-panel executions averaged · 1020/1020 scored units · 100% score-unit run coverage Panel v1 · 95% interval 84.9%–90.6% · generated Jul 9, 2026 Open latest contributing execution logicif-frontier8.8% 12.5% score × 70.0% weight = 8.8% Strict function pass rate · 1 full-panel execution averaged · 24/24 scored units · 100% score-unit run coverage Panel v3 · 95% interval 0.0%–25.0% · generated Jul 29, 2026 Open latest contributing execution	24.8%^† agentic category 24.8%^† Weighted sum of the benchmark contributions below. 24.8%^† category score × 27.5% all-rounder weight = 6.8% bfcl-lite24.8%^† 82.8%^† score × 30.0% weight = 24.8% Mean verifier score · 3 full-panel executions averaged · 180/180 scored units · 100% score-unit run coverage Panel v1 · 95% interval 70.0%–90.0% · generated Jul 9, 2026 Open latest contributing execution thinkingbox-bench0.0% 0.0% score × 70.0% weight = 0.0% Mean task success · 1 full-panel execution averaged · 20/20 scored units · 100% score-unit run coverage Panel v1 · 95% interval 0.0%–0.0% · generated Sep 5, 2026 Open latest contributing execution	555ms	7.77s	~342	149.1s	0.5%	compare
gpt-6-astraMax OpenAI Current leaderboard requirements Every active score-policy benchmark must meet its scored-run and score-unit coverage requirement in this window. gpqa-diamondnot included Failed:0.0% / 95.0% score-unit coverage at 3 scored runs per score unit Needs 95.0% of score units with at least 3 scored runs. Open the model runs to retry availability failures. ifevalnot included Failed:0.0% / 95.0% score-unit coverage at 3 scored runs per score unit Needs 95.0% of score units with at least 3 scored runs. Open the model runs to retry availability failures. logicif-frontierincluded Passed:100% / 100% score-unit coverage at 1 scored runs per score unit humaneval-plusnot included Failed:0.0% / 95.0% score-unit coverage at 3 scored runs per score unit Needs 95.0% of score units with at least 3 scored runs. Open the model runs to retry availability failures. livecodebench-plus-hardnot included Failed:no evidence / 100% score-unit coverage at 1 scored runs per score unit Needs 100% of score units with at least 1 scored runs. Open the model runs to retry availability failures. bfcl-litenot included Failed:0.0% / 95.0% score-unit coverage at 3 scored runs per score unit Needs 95.0% of score units with at least 3 scored runs. Open the model runs to retry availability failures. thinkingbox-benchnot included Failed:no evidence / 100% score-unit coverage at 1 scored runs per score unit Needs 100% of score units with at least 1 scored runs. Open the model runs to retry availability failures.	unqualified Re-weighting of old benchmarks · policy v5 Activated Sep 6, 2026. Every category is required. Open scoring policy intelligenceincomplete Missing required benchmark evidence; no contribution is assigned. 0/1 benchmarks · 0 scored units instructionincomplete Missing required benchmark evidence; no contribution is assigned. 1/2 benchmarks · 24 scored units codingincomplete Missing required benchmark evidence; no contribution is assigned. 0/2 benchmarks · 0 scored units agenticincomplete Missing required benchmark evidence; no contribution is assigned. 0/2 benchmarks · 0 scored units	unqualified Missing 1 weighted intelligence benchmark on its active panel. Every positive-weight benchmark is required for a category score. intelligence category unqualified Weighted sum of the benchmark contributions below. gpqa-diamondmissing Not qualified: 0.0%/95.0% of score units have at least 3 scored runs. Weight: 100%.	unqualified Missing 2 weighted coding benchmarks on its active panel. Every positive-weight benchmark is required for a category score. coding category unqualified Weighted sum of the benchmark contributions below. humaneval-plusmissing Not qualified: 0.0%/95.0% of score units have at least 3 scored runs. Weight: 40.0%. livecodebench-plus-hardmissing No full-panel evidence exists in this window. Weight: 60.0%.	unqualified Missing 1 weighted instruction benchmark on its active panel. Every positive-weight benchmark is required for a category score. instruction category unqualified Weighted sum of the benchmark contributions below. ifevalmissing Not qualified: 0.0%/95.0% of score units have at least 3 scored runs. Weight: 30.0%. logicif-frontier70.0% 100% score × 70.0% weight = 70.0% Strict function pass rate · 1 full-panel execution averaged · 24/24 scored units · 100% score-unit run coverage Panel v3 · 95% interval 100%–100% · generated Jul 29, 2026 Open latest contributing execution	unqualified Missing 2 weighted agentic benchmarks on its active panel. Every positive-weight benchmark is required for a category score. agentic category unqualified Weighted sum of the benchmark contributions below. bfcl-litemissing Not qualified: 0.0%/95.0% of score units have at least 3 scored runs. Weight: 30.0%. thinkingbox-benchmissing No full-panel evidence exists in this window. Weight: 70.0%.	7.31s	18.9s	124	114.4s	0.0%	compare
Meta: Muse Spark 1.3Max OpenRouter Current leaderboard requirements Every active score-policy benchmark must meet its scored-run and score-unit coverage requirement in this window. gpqa-diamondnot included Failed:no evidence / 95.0% score-unit coverage at 3 scored runs per score unit Needs 95.0% of score units with at least 3 scored runs. Open the model runs to retry availability failures. ifevalnot included Failed:no evidence / 95.0% score-unit coverage at 3 scored runs per score unit Needs 95.0% of score units with at least 3 scored runs. Open the model runs to retry availability failures. logicif-frontierincluded Passed:100% / 100% score-unit coverage at 1 scored runs per score unit humaneval-plusnot included Failed:no evidence / 95.0% score-unit coverage at 3 scored runs per score unit Needs 95.0% of score units with at least 3 scored runs. Open the model runs to retry availability failures. livecodebench-plus-hardnot included Failed:no evidence / 100% score-unit coverage at 1 scored runs per score unit Needs 100% of score units with at least 1 scored runs. Open the model runs to retry availability failures. bfcl-liteincluded Passed:100% / 95.0% score-unit coverage at 3 scored runs per score unit thinkingbox-benchincluded Passed:100% / 100% score-unit coverage at 1 scored runs per score unit	unqualified Re-weighting of old benchmarks · policy v5 Activated Sep 6, 2026. Every category is required. Open scoring policy intelligenceincomplete Missing required benchmark evidence; no contribution is assigned. 0/1 benchmarks · 0 scored units instructionincomplete Missing required benchmark evidence; no contribution is assigned. 1/2 benchmarks · 24 scored units codingincomplete Missing required benchmark evidence; no contribution is assigned. 0/2 benchmarks · 0 scored units agentic6.7%^† 24.3%^† score × 27.5% weight = 6.7% 2/2 benchmarks · 80 scored units	unqualified Missing 1 weighted intelligence benchmark on its active panel. Every positive-weight benchmark is required for a category score. intelligence category unqualified Weighted sum of the benchmark contributions below. gpqa-diamondmissing No full-panel evidence exists in this window. Weight: 100%.	unqualified Missing 2 weighted coding benchmarks on its active panel. Every positive-weight benchmark is required for a category score. coding category unqualified Weighted sum of the benchmark contributions below. humaneval-plusmissing No full-panel evidence exists in this window. Weight: 40.0%. livecodebench-plus-hardmissing No full-panel evidence exists in this window. Weight: 60.0%.	unqualified Missing 1 weighted instruction benchmark on its active panel. Every positive-weight benchmark is required for a category score. instruction category unqualified Weighted sum of the benchmark contributions below. ifevalmissing No full-panel evidence exists in this window. Weight: 30.0%. logicif-frontier64.2% 91.7% score × 70.0% weight = 64.2% Strict function pass rate · 1 full-panel execution averaged · 24/24 scored units · 100% score-unit run coverage Panel v3 · 95% interval 79.2%–100% · generated Jul 29, 2026 Open latest contributing execution	24.3%^† agentic category 24.3%^† Weighted sum of the benchmark contributions below. 24.3%^† category score × 27.5% all-rounder weight = 6.7% bfcl-lite24.3%^† 81.1%^† score × 30.0% weight = 24.3% Mean verifier score · 1 full-panel execution averaged · 60/60 scored units · 100% score-unit run coverage Panel v1 · 95% interval 70.6%–90.0% · generated Jul 9, 2026 Open latest contributing execution thinkingbox-bench0.0% 0.0% score × 70.0% weight = 0.0% Mean task success · 1 full-panel execution averaged · 20/20 scored units · 100% score-unit run coverage Panel v1 · 95% interval 0.0%–0.0% · generated Sep 5, 2026 Open latest contributing execution	2.11s	3.26s	~522	12.1s	0.0%	compare

Leaderboard

Category leaders

Best all-rounder

Best intelligence

Best coding

Best instruction

Best agentic