GPT-5.4 Mini places in the top three on 2 benchmarks.
Closed weights Small
Its best score on each benchmark, and where that ranks among every model with a score there. We rank each model once, at its best setting.
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| Global PIQA | 89.6% | #5 of 44 | Official leaderboard | |
| Wer wird Millionär? LLM Benchmark | 225000.0 | #11 of 77 | Official leaderboard | |
| Vectara Hallucination Leaderboard (2025 dataset) | 5.50% | #20 of 104 | Official leaderboard | |
| LMArena (Elo) | 1,447 | #68 of 335 | Official leaderboard | |
| Longform Creative Writing (EQ-Bench) | 73.0 | #33 of 119 | Official leaderboard | |
| HELM Arabic Enterprise | 70.2 | #7 of 23 | Official leaderboard | |
| Judgemark v2.1 | 81.1 | #13 of 42 | Official leaderboard | |
| Creative Writing v3 | 72.5 | #39 of 120 | Official leaderboard | |
| SEA-HELM | 72.8 | #24 of 71 | Official leaderboard | |
| HELM Arabic | 79.0 | #13 of 28 | Official leaderboard | |
| GraphWalks | 76.3% | #14 of 27 | Lab model card | |
| Humanity's Last Exam | 28.2% | #84 of 154 | Lab model card | |
| BullshitBench | 25.4% | #60 of 107 | Official leaderboard | |
| BullshitBench V2 | 32% | #75 of 129 | Official leaderboard | |
| Extended NYT Connections | 61.8% | #64 of 103 | Official leaderboard | |
| Last Translation Benchmark | 2.74 | #24 of 36 | Official leaderboard | |
| OpenAI MRCR v2 | 28.3% | #35 of 47 | Official leaderboard | |
| SimpleQA Verified | 29.4% | #74 of 94 | Epoch AI run | |
| Judgemark v4 | 27.3 | #38 of 40 | Official leaderboard | |
| Antidote | 929 | #35 of 36 | Official leaderboard | |
| MemEye | 49.4% | #1 of 1 | Official leaderboard |
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| FrontierMath | 28.3% | #17 of 55 | Official leaderboard | |
| GPQA Diamond | 86.9% | #73 of 229 | Epoch AI run | |
| OTIS Mock AIME 2024-2025 | 88.9% | #73 of 212 | Epoch AI run | |
| ProofBench | 21% | #29 of 59 | Official leaderboard | |
| DTBench | 80% | #94 of 180 | Official leaderboard | |
| ARC-AGI-2 | 18.9% | #53 of 98 | Official leaderboard | |
| FrontierMath Tiers 1-3 v2 | 51.2% | #53 of 86 | Official leaderboard | |
| ARC-AGI | 63.7% | #60 of 95 | Official leaderboard | |
| BixBench3 | 36.4 | #15 of 20 | Official leaderboard | |
| ForecastBench | 0.15 | #66 of 82 | Official leaderboard | |
| FrontierMath Tier 4 | 2.08% | #45 of 55 | Epoch AI run | |
| FrontierMath Tier 4 v2 | 9.76% | #57 of 68 | Official leaderboard |
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| WeirdML | 60.3% | #35 of 112 | Official leaderboard | |
| FrontierCode | 17.8% | #4 of 12 | Official leaderboard | |
| OpenAI-Proof Q&A | 7.50% | #3 of 9 | Lab model card | |
| Spider 2.0 | 50.5% | #8 of 23 | Official leaderboard | |
| ALE-Bench | 1,487 | #59 of 126 | Official leaderboard | |
| Vibe Code Bench | 48% | #67 of 120 | Official leaderboard | |
| DAXBench v2 | 75.8% | #29 of 47 | Official leaderboard | |
| SWE-Bench Pro | 54.4% | #52 of 81 | Lab model card | |
| WebDev Arena | 1,397 | #68 of 105 | arena.ai, via Epoch AI | |
| Code Arena | 1,398 | #88 of 128 | Official leaderboard | |
| IOI (Vals) | 6.42% | #39 of 55 | Official leaderboard | |
| ProgramBench | 0% | #18 of 25 | Official leaderboard | |
| FrontierCode 1.1 | 27% | #6 of 8 | Official leaderboard | |
| GameDevBench | 43.2% | #13 of 17 | Official leaderboard | |
| Code Migration | 12.9% | #61 of 78 | Official leaderboard | |
| SlopCodeBench | 13.8% | #18 of 22 | Official leaderboard | |
| SWE-sweep | 0.50% | #9 of 10 | Official leaderboard | |
| ProgramBench tests passed | 16.4% | #24 of 25 | Official leaderboard | |
| ProgramBench almost solved | 0% | #25 of 25 | Official leaderboard | |
| ts-bench | 0% | #7 of 7 | Official leaderboard | |
| Monorepo-Bench | 54% | #5 of 5 | Lab model card |
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| tau2-bench | 93.4% | #9 of 79 | Lab model card | |
| Terminal-Bench 2.0 | 60% | #19 of 63 | Lab model card | |
| Toolathlon | 42.9% | #16 of 50 | Official leaderboard | |
| TextQuests | 15.4 | #9 of 28 | Official leaderboard | |
| PinchBench 2.0 | 79.2% | #19 of 57 | Official leaderboard | |
| PinchBench | 71.4% | #18 of 48 | Official leaderboard | |
| LLM Chess | 416 | #47 of 123 | Official leaderboard | |
| LisanBench | 1140.7 | #54 of 133 | Official leaderboard | |
| GAIA | 64.5% | #26 of 64 | Official leaderboard | |
| Game Arena Five in a Row | 497 | #10 of 20 | Official leaderboard | |
| Game Arena Poker | -0.17 | #16 of 31 | Official leaderboard | |
| OSWorld-Verified | 72.1% | #22 of 41 | Lab model card | |
| Chess Puzzles | 23% | #60 of 93 | Epoch AI run | |
| RuneBench | 4.14 | #54 of 76 | Official leaderboard | |
| Game Arena Werewolf | -6 | #6 of 8 | Official leaderboard | |
| Game Arena Four in a Row | 204 | #33 of 43 | Official leaderboard | |
| Game Arena Clobber | 150 | #30 of 39 | Official leaderboard | |
| Game Arena Lines of Action | 274 | #30 of 39 | Official leaderboard | |
| Game Arena Dots and Boxes | 607 | #31 of 40 | Official leaderboard | |
| MyPCBench | 23.9% | #7 of 9 | Official leaderboard | |
| UniClawBench | 21.5% | #8 of 10 | Official leaderboard | |
| o11y-bench | 54.7% | #18 of 22 | Official leaderboard | |
| SkillsBench | 41.4% | #18 of 22 | Official leaderboard | |
| Mystery Game Puzzles | 11% | #69 of 84 | Epoch AI run | |
| ClawBench | 24.8% | #5 of 6 | Official leaderboard | |
| iOSWorld | 26.3% | #5 of 6 | Official leaderboard | |
| Game Arena Word Association | 64 | #34 of 40 | Official leaderboard | |
| Game Arena Dark Hex | 94 | #36 of 41 | Official leaderboard | |
| Game Arena Chess | 438 | #37 of 42 | Official leaderboard | |
| Game Arena Word Art | 126 | #35 of 39 | Official leaderboard | |
| SnakeBench | 15 | #172 of 188 | Official leaderboard | |
| TUA-Bench | 27.2% | #11 of 12 | Official leaderboard | |
| MCP Atlas | 56.7% | #34 of 37 | Official leaderboard | |
| Game Arena Nine Men's Morris | 213 | #36 of 39 | Official leaderboard | |
| Game Arena Reversi | 91 | #36 of 39 | Official leaderboard | |
| Terminal-Bench 4.0 | 1.82% | #37 of 40 | Official leaderboard | |
| Game Arena Checkers | 279 | #38 of 41 | Official leaderboard | |
| Terminal-Bench 4.1 | 1.82% | #40 of 43 | Official leaderboard | |
| Kaggle Game Arena | 143 | #40 of 42 | Official leaderboard | |
| Deep Research Bench (FutureSearch) | 0.36 | #20 of 21 | Official leaderboard | |
| DeepResearch Bench | 36.3% | #21 of 22 | evals.futuresearch.ai, via Epoch AI | |
| Game Arena Coin Game | 89 | #38 of 39 | Official leaderboard | |
| Game Arena Go | 120 | #38 of 39 | Official leaderboard | |
| Game Arena Bargaining | 0 | #39 of 39 | Official leaderboard | |
| Game Arena Ultimate Tic-Tac-Toe | 0 | #39 of 39 | Official leaderboard | |
| Era by Eon | 42.4% | #9 of 9 | Official leaderboard | |
| PlanBench-XL | 0% | #9 of 9 | Official leaderboard |
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| Vals Index v1 | 55.7% | #12 of 34 | Official leaderboard | |
| Vals Finance Agent | 53.4% | #20 of 49 | Official leaderboard | |
| CorpFin v2 | 60.9% | #63 of 119 | Official leaderboard | |
| MortgageTax | 63.5% | #49 of 87 | Official leaderboard | |
| TaxEval v2 | 71.2% | #75 of 129 | Official leaderboard | |
| QF-Bench | 57.1% | #6 of 9 | Official leaderboard | |
| Vals Finance Agent v2 | 45.4% | #55 of 80 | Official leaderboard | |
| Vals Index v2 | 39.6% | #46 of 65 | Official leaderboard | |
| Excel Modeling Benchmark | 45.4% | #58 of 76 | Official leaderboard | |
| 3DCodeArena | 920 | #11 of 14 | Official leaderboard | |
| Tax Agent Bench | 8.81% | #71 of 84 | Official leaderboard | |
| CHI-Bench | 8.40% | #20 of 23 | Official leaderboard | |
| Legal Research Bench | 12.5% | #81 of 93 | Official leaderboard | |
| Harvey LAB (Legal Agent Benchmark) | 0% | #70 of 80 | Official leaderboard | |
| FoodTruck Bench | $470 | #52 of 59 | Official leaderboard | |
| TRLawBench | 7.76 | #30 of 34 | Official leaderboard | |
| CaseLaw v2 | 51.7% | #47 of 53 | Official leaderboard | |
| YC-Bench | $137,649 | #53 of 59 | Official leaderboard | |
| Vals Index v2.1 | 33.2% | #46 of 49 | Official leaderboard | |
| PolicyBench | 73.3% | #47 of 50 | Official leaderboard |
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| GeoGuesser environment (FineEnvs) | 0.45 | #2 of 14 | Official leaderboard | |
| MMMU-Pro | 78% | #28 of 90 | Lab model card | |
| Vision Arena | 1,250 | #44 of 136 | Official leaderboard | |
| ZeroBench | 11% | #37 of 76 | Official leaderboard | |
| MineBench | 1,387 | #45 of 75 | Official leaderboard | |
| BVB | 60.2 | #36 of 45 | Official leaderboard | |
| vCX Benchmark | 77.9% | #37 of 42 | Official leaderboard | |
| AgenticVBench | 14.5% | #9 of 9 | Official leaderboard | |
| OmniDocBench v1.5 | 0.13 | #9 of 9 | Lab model card |
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| Capture the Flag (Professional) | 81.3% | #7 of 21 | Lab model card | |
| nocot-bench (NCRI) | 70.5 | #98 of 185 | Official leaderboard | |
| Tacit Knowledge and Troubleshooting | 71.7% | #17 of 30 | Lab model card | |
| CVE-Bench | 83.3% | #8 of 14 | Lab model card | |
| ProtocolQA Open-Ended | 33.6% | #17 of 27 | Lab model card | |
| TroubleshootingBench | 31.9% | #15 of 20 | Lab model card | |
| CRSBench | 54.9% | #4 of 5 | Official leaderboard | |
| Multimodal Troubleshooting Virology | 46.5% | #23 of 27 | Lab model card | |
| PCT (Polarized Contrastive Pairs) | 61.1% | #44 of 46 | Official leaderboard |
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| SAGE (Vals) | 50.8% | #15 of 81 | Official leaderboard | |
| ACCoRD | 70.5 | #52 of 152 | Official leaderboard | |
| LMCA | 40.8% | #55 of 147 | Official leaderboard | |
| Conceptual Reasoning Index | 51.9 | #63 of 151 | Official leaderboard | |
| StudentBench | 13.2 | #6 of 11 | Official leaderboard | |
| CAIS Text Capabilities Index | 26.6 | #40 of 68 | Official leaderboard | |
| CAIS Risk Index | 53.9 | #40 of 67 | Official leaderboard | |
| Vals Multimodal Index | 54.2% | #21 of 33 | Official leaderboard | |
| CAIS Vision Capabilities Index | 46.6 | #34 of 53 | Official leaderboard |
The Self-Reported Index fits one score to the 12 capability results OpenAI reported for this model, on the scale of the Epoch Capabilities Index. Labs choose which benchmarks to show and how to run them, so read it as the lab's claim, not an independent measurement. The 90% interval runs from 144.8 to 150.5. How we compute it.
Where the benchmark's own leaderboard, or Epoch AI's run, also has this model. Gaps mostly come from different harnesses, effort levels and task subsets. Every pair.
| Benchmark | Lab says | Board says | Gap |
|---|---|---|---|
| OpenAI MRCR v2 | 47.7% OpenAI MRCR v2 8-needle, 64K-128K | 28.3% Leaderboard, 8-needle, 128K to 256K bin, high reasoning | +19 pts |
| GPQA Diamond | 88% GPQA Diamond | 86.9% Epoch AI run, xhigh | +1.1 pts |
| MCP Atlas | 57.7% MCP-Atlas | 56.7% Leaderboard, xhigh | +1.0 pts |
| Toolathlon | 42.9% Toolathlon | 42.9% Leaderboard | Same |
OpenAI's models in color, against every other model on the Epoch Capabilities Index, by release date.
Source: Epoch Capabilities Index, Epoch AI (CC BY 4.0)
The benchmarks the lab chose to headline, with the scores it gave.
Every result in the longer documents the lab published with it. 11 of them aren't in the launch post.