Consensus score
83.5
95% interval 74.5 to 90.2
83.5
95% interval 74.5 to 90.2
86th
IQR spread 18.2 pct
108
Across 9 domains
4
Newest observation 2d ago
Median within-board percentile by capability domain across boards with sufficient peer overlap.
Best verified run per board, with confidence intervals and Pareto frontier status.
| Benchmark | Rank | Score | Cost | Status | Observed |
|---|---|---|---|---|---|
| ARC-AGI-2 Epoch AI Benchmarking Hub · max | #1 / 299Tied #1 | 95% | Unknown | Standard | Not stated |
| HLE Epoch AI Benchmarking Hub · Default | #1 / 56Tied #1 | 54.8%[51% to 58.6%] | Unknown | Standard | Not stated |
| LMArena WebDev LMArena · Webdev HTML · max | #2 / 155Tied #1 | 1,786[1,767 to 1,804] | Unknown | Standard | 15 Nov 2026 |
| Vending-Bench 2 Epoch AI Benchmarking Hub · Default | #2 / 72 | $15,514.70 | Unknown | Standard | Not stated |
| LMArena Search (style control) LMArena · Overall · max | #3 / 42 | 1,245[1,237 to 1,253] | Unknown | Standard | 9 Nov 2026 |
| FrontierCode Epoch AI Benchmarking Hub · max · Codex CLI | #4 / 51 | 53.3% | Unknown | Standard | Not stated |
| GPQA diamond Epoch AI Benchmarking Hub · max | #4 / 338Tied #1 | 95.8%[93.1% to 98.5%] | Unknown | Standard | 30 Aug 2026 |
| LMArena WebDev LMArena · Image to webdev · max | #4 / 64 | 1,719[1,708 to 1,730] | Unknown | Standard | 15 Nov 2026 |
| ARC-AGI-2 Epoch AI Benchmarking Hub · xhigh | #5 / 299 | 93.3% | Unknown | Standard | Not stated |
| LMArena Text (style control) LMArena · Coding · max | #5 / 416Tied #1 | 1,548[1,538 to 1,558] | Unknown | Standard | 11 Nov 2026 |
| LMArena Vision (style control) LMArena · Homework · max | #5 / 118Tied #1 | 1,337[1,316 to 1,358] | Unknown | Standard | 9 Nov 2026 |
| LMArena WebDev LMArena · Overall · max | #5 / 156 | 1,763[1,754 to 1,771] | Unknown | Standard | 15 Nov 2026 |
| LMArena WebDev LMArena · Webdev · max | #5 / 156 | 1,763[1,754 to 1,771] | Unknown | Standard | 15 Nov 2026 |
| Code Migration Vals AI benchmarks · max | #6 / 38 | 78%[71.9% to 84.2%] | $66.64 | Standard | 15 Nov 2026 |
| LMArena WebDev LMArena · Webdev react · max | #6 / 140 | 1,775[1,764 to 1,785] | Unknown | Standard | 15 Nov 2026 |
| Terminal-Bench 4 Vals AI benchmarks · max | #6 / 38Tied #1 | 59.6%[51% to 68.2%] | $9.58 | Standard | 15 Nov 2026 |
| Vibe Code Bench Vals AI benchmarks · max · OpenHands | #6 / 38Tied #1 | 89.6%[85.3% to 93.9%] | $38.51 | Standard | 15 Nov 2026 |
| Bug Hunt Bench Bug Hunt Bench · max · Codex CLI | #7 / 145 | 45 | $33.03 | Standard | 14 Sept 2026 |
| LMArena Agent LMArena · Overall · max | #7 / 64 | 0.1[0.1 to 0.2] | $5.28 | Standard | 15 Nov 2026 |
| LMArena Search LMArena · Overall · max | #7 / 42 | 1,231[1,223 to 1,239] | Unknown | Standard | 9 Nov 2026 |
| Vals Index Vals AI benchmarks · max | #8 / 38 | 65%[62.9% to 67.2%] | $21.28 | Standard | 15 Nov 2026 |
| LMArena Vision (style control) LMArena · English · max | #9 / 167Tied #1 | 1,289[1,274 to 1,304] | Unknown | Standard | 9 Nov 2026 |
| ARC-AGI-2 Epoch AI Benchmarking Hub · high | #10 / 299 | 92.1% | Unknown | Standard | Not stated |
| ARC-AGI-2 Epoch AI Benchmarking Hub · medium | #10 / 299 | 92.1% | Unknown | Standard | Not stated |
| Bug Hunt Bench Bug Hunt Bench · xhigh · Codex CLI | #10 / 145 | 43 | $24.22 | Standard | 4 Sept 2026 |
| EMB Vals AI benchmarks · max | #10 / 38Tied #1 | 72.7%[68.2% to 77.2%] | $5.83 | Standard | 15 Nov 2026 |
| LMArena Text (style control) LMArena · German · max | #12 / 320 | 1,500[1,465 to 1,535] | Unknown | Standard | 11 Nov 2026 |
| LMArena Vision LMArena · Homework · max | #13 / 118 | 1,332[1,311 to 1,353] | Unknown | Standard | 9 Nov 2026 |
| LMArena Text (style control) LMArena · Industry software and IT services · max | #14 / 421 | 1,526[1,518 to 1,534] | Unknown | Standard | 11 Nov 2026 |
| LMArena Vision (style control) LMArena · Creative writing vision · max | #14 / 106 | 1,289[1,262 to 1,316] | Unknown | Standard | 9 Nov 2026 |
| LMArena Document (style control) LMArena · Overall · max | #17 / 46 | 1,474[1,465 to 1,483] | Unknown | Standard | 9 Nov 2026 |
| LMArena Vision (style control) LMArena · Humor · max | #17 / 102 | 1,282[1,242 to 1,323] | Unknown | Standard | 9 Nov 2026 |
| LMArena Document LMArena · Overall · max | #18 / 46 | 1,467[1,459 to 1,476] | Unknown | Standard | 9 Nov 2026 |
| LMArena Vision (style control) LMArena · Overall · max | #21 / 167 | 1,283[1,274 to 1,293] | Unknown | Standard | 9 Nov 2026 |
| LMArena Text (style control) LMArena · Hard prompts English · max | #23 / 419 | 1,510[1,501 to 1,520] | Unknown | Standard | 11 Nov 2026 |
| LMArena Text (style control) LMArena · Expert · max | #24 / 371 | 1,517[1,502 to 1,531] | Unknown | Standard | 11 Nov 2026 |
| LMArena Text (style control) LMArena · Industry mathematical · max | #24 / 402 | 1,499[1,480 to 1,517] | Unknown | Standard | 11 Nov 2026 |
| LMArena Vision (style control) LMArena · Diagram · max | #24 / 123 | 1,311[1,294 to 1,328] | Unknown | Standard | 9 Nov 2026 |
| LMArena Vision (style control) LMArena · OCR · max | #24 / 123 | 1,297[1,287 to 1,307] | Unknown | Standard | 9 Nov 2026 |
| Legal Research Vals AI benchmarks · max | #24 / 38 | 39.4%[32.8% to 46.1%] | $10.58 | Standard | 15 Nov 2026 |
| Finance Agent Vals AI benchmarks · max | #26 / 38 | 60.4%[56.3% to 64.6%] | $7.82 | Standard | 15 Nov 2026 |
| LMArena Text (style control) LMArena · Industry entertainment and sports and media · max | #27 / 419 | 1,456[1,446 to 1,465] | Unknown | Standard | 11 Nov 2026 |
| LMArena Text (style control) LMArena · Korean · max | #28 / 289Tied #1 | 1,446[1,409 to 1,483] | Unknown | Standard | 11 Nov 2026 |
| Legal Agent Benchmark (HLAB-Vals) Vals AI benchmarks · max | #28 / 38 | 7.1%[3.4% to 10.8%] | $25.46 | Standard | 15 Nov 2026 |
| LMArena Text (style control) LMArena · English · max | #29 / 421 | 1,487[1,479 to 1,494] | Unknown | Standard | 11 Nov 2026 |
| LMArena Vision (style control) LMArena · Chinese · max | #29 / 135 | 1,327[1,294 to 1,361] | Unknown | Standard | 9 Nov 2026 |
| Tax Agent Vals AI benchmarks · max | #31 / 38 | 20.7%[15% to 26.4%] | $5.80 | Standard | 15 Nov 2026 |
| LMArena Text (style control) LMArena · Industry business and management and financial operations · max | #32 / 414 | 1,478[1,467 to 1,489] | Unknown | Standard | 11 Nov 2026 |
| LMArena Vision LMArena · Creative writing vision · max | #32 / 106 | 1,278[1,251 to 1,305] | Unknown | Standard | 9 Nov 2026 |
| LMArena Text LMArena · Coding · max | #34 / 416 | 1,494[1,484 to 1,504] | Unknown | Standard | 11 Nov 2026 |
| LMArena Text (style control) LMArena · Instruction following · max | #34 / 421 | 1,474[1,465 to 1,482] | Unknown | Standard | 11 Nov 2026 |
| LMArena Text (style control) LMArena · Exclude ties · max | #35 / 421 | 1,486[1,479 to 1,493] | Unknown | Standard | 11 Nov 2026 |
| LMArena Text (style control) LMArena · Industry writing and literature and language · max | #35 / 420 | 1,465[1,455 to 1,474] | Unknown | Standard | 11 Nov 2026 |
| LMArena Vision LMArena · Diagram · max | #35 / 123 | 1,299[1,282 to 1,316] | Unknown | Standard | 9 Nov 2026 |
| Bug Hunt Bench Bug Hunt Bench · high · Codex CLI | #36 / 145 | 35 | $20.60 | Standard | 4 Sept 2026 |
| LMArena Text LMArena · Expert · max | #36 / 371 | 1,493[1,478 to 1,507] | Unknown | Standard | 11 Nov 2026 |
| LMArena Text (style control) LMArena · Chinese · max | #36 / 395 | 1,520[1,504 to 1,536] | Unknown | Standard | 11 Nov 2026 |
| LMArena Text (style control) LMArena · Hard prompts · max | #36 / 421 | 1,500[1,493 to 1,506] | Unknown | Standard | 11 Nov 2026 |
| LMArena Text (style control) LMArena · Multi turn · max | #36 / 419 | 1,483[1,471 to 1,495] | Unknown | Standard | 11 Nov 2026 |
| LMArena Vision LMArena · Chinese · max | #36 / 135 | 1,332[1,299 to 1,365] | Unknown | Standard | 9 Nov 2026 |
| LMArena Vision LMArena · English · max | #36 / 167 | 1,282[1,267 to 1,297] | Unknown | Standard | 9 Nov 2026 |
| LMArena Vision LMArena · OCR · max | #36 / 123 | 1,290[1,279 to 1,300] | Unknown | Standard | 9 Nov 2026 |
| LMArena Text (style control) LMArena · Industry legal and government · max | #37 / 393 | 1,482[1,466 to 1,498] | Unknown | Standard | 11 Nov 2026 |
| ARC-AGI-2 Epoch AI Benchmarking Hub · low | #38 / 299 | 85.4% | Unknown | Standard | Not stated |
| LMArena Text LMArena · German · max | #38 / 320 | 1,465[1,430 to 1,500] | Unknown | Standard | 11 Nov 2026 |
| LMArena Text (style control) LMArena · Longer query · max | #38 / 399 | 1,485[1,478 to 1,493] | Unknown | Standard | 11 Nov 2026 |
| LMArena Text (style control) LMArena · Math · max | #38 / 404 | 1,481[1,462 to 1,500] | Unknown | Standard | 11 Nov 2026 |
| LMArena Vision LMArena · Humor · max | #38 / 102 | 1,272[1,232 to 1,312] | Unknown | Standard | 9 Nov 2026 |
| LMArena Vision LMArena · Overall · max | #38 / 167 | 1,279[1,269 to 1,288] | Unknown | Standard | 9 Nov 2026 |
| Bug Hunt Bench Bug Hunt Bench · medium · Codex CLI | #40 / 145 | 34 | $15.78 | Standard | 5 Sept 2026 |
| LMArena Text (style control) LMArena · Creative writing · max | #40 / 419 | 1,451[1,440 to 1,461] | Unknown | Standard | 11 Nov 2026 |
| LMArena Text LMArena · Korean · max | #42 / 289 | 1,424[1,387 to 1,460] | Unknown | Standard | 11 Nov 2026 |
| LMArena Text (style control) LMArena · Overall · max | #42 / 421 | 1,474[1,468 to 1,479] | Unknown | Standard | 11 Nov 2026 |
| LMArena Text (style control) LMArena · Spanish · max | #44 / 304 | 1,466[1,442 to 1,490] | Unknown | Standard | 11 Nov 2026 |
| LMArena Text LMArena · Industry mathematical · max | #45 / 402 | 1,478[1,459 to 1,497] | Unknown | Standard | 11 Nov 2026 |
| LMArena Text (style control) LMArena · Industry life and physical and social science · max | #45 / 419 | 1,488[1,477 to 1,500] | Unknown | Standard | 11 Nov 2026 |
| LMArena Text (style control) LMArena · Japanese · max | #45 / 288 | 1,446[1,412 to 1,480] | Unknown | Standard | 11 Nov 2026 |
| LMArena Text (style control) LMArena · Non English · max | #45 / 421 | 1,461[1,454 to 1,467] | Unknown | Standard | 11 Nov 2026 |
| LMArena Text (style control) LMArena · French · max | #46 / 301 | 1,486[1,460 to 1,512] | Unknown | Standard | 11 Nov 2026 |
| LMArena Text (style control) LMArena · Russian · max | #48 / 385 | 1,471[1,459 to 1,483] | Unknown | Standard | 11 Nov 2026 |
| LMArena Text LMArena · Instruction following · max | #49 / 421 | 1,456[1,447 to 1,464] | Unknown | Standard | 11 Nov 2026 |
| LMArena Text LMArena · Industry software and IT services · max | #52 / 421 | 1,479[1,471 to 1,487] | Unknown | Standard | 11 Nov 2026 |
| LMArena Text (style control) LMArena · Polish · max | #53 / 243 | 1,466[1,433 to 1,500] | Unknown | Standard | 11 Nov 2026 |
| LMArena Text LMArena · Hard prompts English · max | #54 / 419 | 1,468[1,458 to 1,478] | Unknown | Standard | 11 Nov 2026 |
| LMArena Text LMArena · Longer query · max | #54 / 399 | 1,459[1,451 to 1,466] | Unknown | Standard | 11 Nov 2026 |
| LMArena Text LMArena · Hard prompts · max | #55 / 421 | 1,462[1,456 to 1,469] | Unknown | Standard | 11 Nov 2026 |
| LMArena Text LMArena · Industry business and management and financial operations · max | #55 / 414 | 1,442[1,431 to 1,453] | Unknown | Standard | 11 Nov 2026 |
| LMArena Text LMArena · Industry entertainment and sports and media · max | #57 / 419 | 1,425[1,415 to 1,434] | Unknown | Standard | 11 Nov 2026 |
| LMArena Text LMArena · Industry writing and literature and language · max | #58 / 420 | 1,438[1,429 to 1,447] | Unknown | Standard | 11 Nov 2026 |
| LMArena Text LMArena · Chinese · max | #60 / 395 | 1,496[1,480 to 1,511] | Unknown | Standard | 11 Nov 2026 |
| LMArena Text LMArena · Math · max | #60 / 404 | 1,459[1,440 to 1,478] | Unknown | Standard | 11 Nov 2026 |
| Bug Hunt Bench Bug Hunt Bench · low · Codex CLI | #62 / 145 | 27 | $11.69 | Standard | 5 Sept 2026 |
| LMArena Text LMArena · Japanese · max | #64 / 288 | 1,415[1,381 to 1,448] | Unknown | Standard | 11 Nov 2026 |
| LMArena Text LMArena · Industry legal and government · max | #66 / 393 | 1,453[1,437 to 1,469] | Unknown | Standard | 11 Nov 2026 |
| LMArena Text LMArena · Russian · max | #67 / 385 | 1,441[1,429 to 1,454] | Unknown | Standard | 11 Nov 2026 |
| LMArena Text LMArena · Creative writing · max | #70 / 419 | 1,421[1,411 to 1,432] | Unknown | Standard | 11 Nov 2026 |
| LMArena Text LMArena · Multi turn · max | #73 / 419 | 1,448[1,436 to 1,460] | Unknown | Standard | 11 Nov 2026 |
| LMArena Text LMArena · Non English · max | #75 / 421 | 1,430[1,423 to 1,436] | Unknown | Standard | 11 Nov 2026 |
| LMArena Text LMArena · Overall · max | #78 / 421 | 1,440[1,434 to 1,445] | Unknown | Standard | 11 Nov 2026 |
| LMArena Text LMArena · English · max | #79 / 421 | 1,448[1,441 to 1,456] | Unknown | Standard | 11 Nov 2026 |
| LMArena Text LMArena · Exclude ties · max | #79 / 421 | 1,435[1,428 to 1,442] | Unknown | Standard | 11 Nov 2026 |
| LMArena Text (style control) LMArena · Industry medicine and healthcare · max | #83 / 391 | 1,468[1,451 to 1,485] | Unknown | Standard | 11 Nov 2026 |
| LMArena Text LMArena · Industry life and physical and social science · max | #88 / 419 | 1,447[1,435 to 1,458] | Unknown | Standard | 11 Nov 2026 |
| LMArena Text LMArena · Spanish · max | #92 / 304 | 1,436[1,411 to 1,460] | Unknown | Standard | 11 Nov 2026 |
| LMArena Text LMArena · Polish · max | #97 / 243 | 1,426[1,392 to 1,460] | Unknown | Standard | 11 Nov 2026 |
| LMArena Text LMArena · French · max | #102 / 301 | 1,440[1,414 to 1,466] | Unknown | Standard | 11 Nov 2026 |
| ARC-AGI-2 Epoch AI Benchmarking Hub · none | #106 / 299 | 59.6% | Unknown | Standard | Not stated |
| LMArena Text LMArena · Industry medicine and healthcare · max | #132 / 391 | 1,420[1,403 to 1,437] | Unknown | Standard | 11 Nov 2026 |
Same benchmark and harness measured across multiple reasoning effort settings.
Epoch AI Benchmarking Hub
59.6
85.4
92.1
92.1
93.3
95.0
Bug Hunt Bench · Codex CLI
27.0
$11.69
34.0
$15.78
35.0
$20.60
43.0
$24.22
45.0
$33.03
Direct pairwise win rates for GPT-6 Astra across shared benchmark boards.
23 shared boards
61% win
Head-to-head22 shared boards
68% win
Head-to-head22 shared boards
27% win
Head-to-head21 shared boards
100% win
Head-to-head21 shared boards
33% win
Head-to-head20 shared boards
80% win
Head-to-head19 shared boards
63% win
Head-to-head19 shared boards
79% win
Head-to-head18 shared boards
44% win
Head-to-head18 shared boards
72% win
Head-to-head