Introducing GPT-6.1 Terra (OpenAI) Oct 26, 2026
GPT-6.1 Terra
GPT-6.1 Terra ranks first on 4 of the 144 benchmarks with five or more models scored.
149
Benchmarks with a score
4
Ranked first, out of five or more models
20
In the top three
OpenAI Oct 2026
GPT-6.1 Terra
Closed weights
167.7
#6 of 256
Epoch Capabilities Index
Self-Reported Index 159.8 #32 of 404
- MultiChallenge 74.2% #7 /42
- Terminal-Bench 4.0 51.8% #8 /40
- GPQA Diamond 94.3% #11 /229
Where it ranks
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.
Knowledge and language
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| FACTS Grounding v2 | 84.2% | #1 of 25 | Official leaderboard | |
| MultiChallenge | 74.2% | #7 of 42 | Official leaderboard | |
| FACTS Benchmark Suite | 74.1% | #4 of 23 | Official leaderboard | |
| SimpleQA Verified | 57.3% | #22 of 94 | Epoch AI run | |
| Extended NYT Connections | 86.8% | #32 of 103 | Official leaderboard | |
| BullshitBench V2 | 54% | #47 of 129 | Official leaderboard | |
| FACTS Parametric V2 | 70.8% | #11 of 27 | Official leaderboard | |
| FACTS Multimodal | 48.8% | #12 of 23 | Official leaderboard |
Math and science
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| GPQA Diamond | 94.3% | #11 of 229 | Epoch AI run | |
| FrontierMath Tiers 1-3 v2 | 90.9% | #5 of 86 | Official leaderboard | |
| OTIS Mock AIME 2024-2025 | 100% | #14 of 212 | Epoch AI run | |
| FrontierMath Tier 4 v2 | 92.7% | #6 of 68 | Official leaderboard | |
| SimpleBench | 72.2% | #22 of 106 | Official leaderboard | |
| FrontierPhysics | 14.3% | #5 of 22 | Official leaderboard | |
| LatchBio Capabilities v1.0 | 43.4% | #12 of 27 | Official leaderboard | |
| scBench (LatchBio SingleCellBench) | 58.3% | #12 of 26 | Official leaderboard | |
| Prophet Arena | 85.7 | #12 of 24 | Official leaderboard |
Coding
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| ALE-Bench | 2,875 | #5 of 126 | Official leaderboard | |
| DAXBench v2 | 91.9% | #3 of 47 | Official leaderboard | |
| Code Arena | 1,679 | #12 of 128 | Official leaderboard | |
| OpenDesign Arena | 90.3 | #2 of 21 | Official leaderboard | |
| DesignArena Game Dev | 1,333 | #14 of 143 | Official leaderboard | |
| Brokk Power Ranking | 29.6 | #2 of 19 | Official leaderboard | |
| MacroscopeBench | 82.3% | #2 of 16 | Official leaderboard | |
| DesignArena SVG | 1,294 | #13 of 103 | Official leaderboard | |
| MirrorCode | 76.7% | #6 of 39 | Epoch AI run | |
| DesignArena Code | 1,300 | #23 of 143 | Official leaderboard | |
| DesignArena 3D Design | 1,305 | #23 of 136 | Official leaderboard | |
| DesignArena Website | 1,287 | #27 of 151 | Official leaderboard | |
| Vibe Code Bench | 84% | #22 of 120 | Official leaderboard | |
| WeirdML v3 | 28.1% | #5 of 27 | Official leaderboard | |
| DesignArena UI Component | 1,289 | #28 of 139 | Official leaderboard | |
| Android Bench 2.0 | 28.7% | #4 of 17 | Official leaderboard | |
| DesignArena Image to HTML | 1,227 | #16 of 51 | Official leaderboard | |
| FrontierCS 2 | 39.6% | #6 of 19 | Official leaderboard | |
| Internal Research Debugging Evaluation | 76.4% | #2 of 6 | Lab model card | |
| Next.js Agent Evals | 100% | #5 of 14 | Official leaderboard | |
| DesignArena Agentic Overall | 1,193 | #30 of 74 | Official leaderboard | |
| DesignArena React Native | 1,180 | #28 of 56 | Official leaderboard | |
| KernelGen 1P | 45.3% | #2 of 4 | Lab model card | |
| SWE Atlas v1.1 | 57.7% | #18 of 28 | Official leaderboard | |
| DesignArena Data Visualization | 1,166 | #81 of 118 | Official leaderboard | |
| Floorplan Arena | 1,403 | #8 of 11 | Official leaderboard |
Agents and tools
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| GameWorld | 76.8% | #1 of 22 | Official leaderboard | |
| Mystery Game Puzzles | 74% | #4 of 84 | Epoch AI run | |
| LLM Chess | 1,521 | #9 of 123 | Official leaderboard | |
| RuneBench | 6.65 | #6 of 76 | Official leaderboard | |
| EBR-bench | 78.1% | #3 of 31 | Epoch AI run | |
| DuelLab GameBench 2 | 85.8 | #8 of 57 | Official leaderboard | |
| Agent Arena | 12% | #8 of 56 | Official leaderboard | |
| Game Arena Word Art | 1,596 | #6 of 39 | Official leaderboard | |
| BU Bench V2.1 | 79.0 | #3 of 16 | Official leaderboard | |
| Terminal-Bench 4.0 | 51.8% | #8 of 40 | Official leaderboard | |
| SpireClimb StS2 A10 | 34.7% | #2 of 9 | Official leaderboard | |
| Terminal-Bench 4.1 | 53.3% | #10 of 43 | Official leaderboard | |
| Game Arena Dark Hex | 348 | #10 of 41 | Official leaderboard | |
| Game Arena Go | 1,202 | #10 of 39 | Official leaderboard | |
| Game Arena Dots and Boxes | 983 | #11 of 40 | Official leaderboard | |
| Kaggle Game Arena | 372 | #12 of 42 | Official leaderboard | |
| Game Arena Word Association | 310 | #13 of 40 | Official leaderboard | |
| Game Arena Chess | 1,069 | #14 of 42 | Official leaderboard | |
| Game Arena Clobber | 411 | #13 of 39 | Official leaderboard | |
| SpireClimb | 22.3% | #3 of 9 | Epoch AI run | |
| Game Arena Checkers | 1,024 | #14 of 41 | Official leaderboard | |
| Game Arena Bargaining | 214 | #14 of 39 | Official leaderboard | |
| Game Arena Ultimate Tic-Tac-Toe | 361 | #17 of 39 | Official leaderboard | |
| SpireClimb StS1 A20 Heart | 10% | #4 of 9 | Official leaderboard | |
| Game Arena Nine Men's Morris | 665 | #18 of 39 | Official leaderboard | |
| Game Arena Reversi | 552 | #19 of 39 | Official leaderboard | |
| Game Arena Lines of Action | 393 | #20 of 39 | Official leaderboard | |
| Game Arena Four in a Row | 333 | #25 of 43 | Official leaderboard | |
| Game Arena Coin Game | 330 | #23 of 39 | Official leaderboard |
Professional work
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| PolicyBench | 93.1% | #3 of 50 | Official leaderboard | |
| APEX-Agents | 68.3% | #8 of 61 | Collected by Epoch AI, source not named | |
| APEX-Agents 1.1 | 79.3% | #10 of 65 | Official leaderboard | |
| Vending-Bench 2 | $9,107 | #14 of 67 | Official leaderboard | |
| Excel Modeling Benchmark v1.1 | 73.3% | #9 of 42 | Official leaderboard | |
| DeepFinance | 47.2% | #6 of 23 | Official leaderboard | |
| APEX-Accounting | 63.7% | #13 of 49 | Official leaderboard | |
| LifeSci Diamond | 9.10% | #6 of 18 | Official leaderboard | |
| TaxCalcBench v2 | 62% | #11 of 32 | Official leaderboard | |
| GDPx | 8.48% | #7 of 20 | Official leaderboard | |
| Vals Index v2.2 | 60.5% | #13 of 37 | Official leaderboard | |
| Agents' Last Exam v1.1 | 41.8% | #5 of 14 | Official leaderboard | |
| HealthBench Professional | 62.7% | #10 of 26 | Lab model card | |
| RWE-bench | 19% | #6 of 15 | Official leaderboard | |
| YC-Bench | $1,154,352 | #29 of 59 | Official leaderboard | |
| APEX-1 | 64.2% | #36 of 71 | Official leaderboard | |
| DesignArena Spreadsheet | 1,146 | #18 of 35 | Official leaderboard | |
| Legal Research Bench | 26% | #64 of 93 | Official leaderboard | |
| HealthBench Consensus | 95.1% | #12 of 17 | Lab model card | |
| Tax Agent Bench | 17.1% | #60 of 84 | Official leaderboard | |
| HealthBench Hard | 31.4% | #16 of 21 | Lab model card | |
| Vals Finance Index | 53.3% | #31 of 40 | Official leaderboard | |
| Company Agent Bench | 76.8% | #5 of 6 | Official leaderboard | |
| HealthBench | 55.1% | #26 of 31 | Lab model card | |
| Vals Finance Agent v2.1 | 56% | #34 of 40 | Official leaderboard | |
| Vals Legal Index | 58.8% | #35 of 38 | Official leaderboard | |
| DesignArena PDF | 1,082 | #35 of 37 | Official leaderboard | |
| Harvey LAB v1.1 (HLAB-Vals-1.1) | 1.67% | #38 of 38 | Official leaderboard |
Vision, audio and robotics
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| BVB | 75.2 | #4 of 45 | Official leaderboard | |
| VoxelBench | 2,314 | #5 of 56 | Official leaderboard | |
| vCX Benchmark | 92.4% | #4 of 42 | Official leaderboard | |
| ParseBench | 73.6% | #5 of 44 | Official leaderboard | |
| Furniture Assembly Benchmark | 79.2% | #5 of 35 | Official leaderboard | |
| ExtractBench | 94.3% | #4 of 27 | Official leaderboard | |
| Blueprint-Bench 2 | 42.1% | #7 of 38 | Official leaderboard | |
| ShadowBench | 46.4 | #3 of 6 | Official leaderboard | |
| DesignArena Code to Video | 1,112 | #19 of 29 | Official leaderboard | |
| ExtractBench Pro | 27% | #5 of 7 | Official leaderboard | |
| RoboWorld | 0.00 | #3 of 3 | Official leaderboard |
Safety and security
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| nocot-bench (NCRI) | 130.5 | #4 of 185 | Official leaderboard | |
| SpatialBench-Long | 40.5% | #2 of 27 | Official leaderboard | |
| Coronavirus-ACE2 Cell-Entry Screen | 0.47 | #1 of 6 | Lab model card | |
| ExploitBench | 97.1% | #4 of 21 | Lab model card | |
| Blocker Deception | 8.47% | #1 of 5 | Lab model card | |
| BioTIER Refuse | 79.1% | #19 of 69 | Official leaderboard | |
| Broken Search Tool | 1.83% | #3 of 9 | Lab model card | |
| Phage-plasmid Co-evolution | 10.8 | #2 of 6 | Lab model card | |
| SEC-bench Pro | 83.3% | #4 of 11 | Lab model card | |
| FORTRESS | 18.2% | #27 of 73 | Official leaderboard | |
| ExploitBench - Internal Port | 24.5% | #2 of 5 | Lab model card | |
| TxBench-Preclinical-Pharmacology | 55.7% | #14 of 30 | Official leaderboard | |
| BioSecBench-Function | 39% | #13 of 27 | Official leaderboard | |
| EpiBench | 23.3% | #13 of 27 | Official leaderboard | |
| TxBench-Antibody-Discovery | 40.7% | #15 of 30 | Official leaderboard | |
| MetagenomicsBench | 38.7% | #13 of 26 | Official leaderboard | |
| SHP2 Protein Function Prediction | 0.37 | #3 of 6 | Lab model card | |
| MentalHealthBench | 55.5% | #2 of 4 | Lab model card | |
| Sandbox Bench | 22.7% | #2 of 4 | Lab model card | |
| Multimodal Troubleshooting Virology | 52.5% | #14 of 27 | Lab model card | |
| scBench-Long | 18.8% | #14 of 27 | Official leaderboard | |
| BioSecBench-Refusal | 34.2% | #12 of 22 | Official leaderboard | |
| VariantBench | 34.8% | #15 of 27 | Official leaderboard | |
| OncologyBench | 16% | #18 of 30 | Official leaderboard | |
| CoT-Control | 14.8% | #6 of 9 | Lab model card | |
| AAV Capsid Packaging Prediction (OpenAI) | 0.53 | #5 of 7 | Lab model card | |
| Respecting Auto-Review | 0% | #5 of 7 | Lab model card | |
| BioTIER Permit | 98.6% | #50 of 69 | Official leaderboard | |
| BioSecBench-Surveillance | 39.9% | #20 of 27 | Official leaderboard | |
| Misaligned Behavior in Realistic Work Environments | 2.28% | #3 of 4 | Lab model card | |
| TxBench-Oligo-Discovery | 30.4% | #24 of 30 | Official leaderboard | |
| OpenAI Deployment Simulation | 0.04% | #4 of 5 | Lab model card | |
| DistressBench | 65.9% | #35 of 42 | Official leaderboard | |
| Declining to Exploit a Honeypot During Difficult ExploitGym Problems | 0.07% | #6 of 7 | Lab model card | |
| Respecting Warnings | 29.6% | #6 of 7 | Lab model card | |
| Biopharma Bench | 40.4% | #13 of 13 | Official leaderboard | |
| Unintended Engagement with External Agent Messages | 65% | #6 of 6 | Lab model card |
Other
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| LiveBench | 78.8% | #7 of 31 | Official leaderboard |
Its lab's numbers put it at 159.8, 7.9 points below its ECI
The Self-Reported Index fits one score to the 5 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 154.0 to 165.5. How we compute it.
- HealthBench Consensus 95.1
- HealthBench Professional 62.7
- HealthBench 55.1
- Terminal-Bench 4.0 50.3
- HealthBench Hard 31.4
Its lab's number next to a leaderboard
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 |
|---|---|---|---|
| Terminal-Bench 4.0 | 50.3% Terminal-Bench 4.0, maximum reasoning effort | 51.8% Leaderboard, max, Codex | −1.5 pts |
OpenAI's models over time
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)
Its launch post
The benchmarks the lab chose to headline, with the scores it gave.
Its system card
Every result in the longer documents the lab published with it. 36 of them aren't in the launch post.
Addendum to GPT-6 Astra System Card: GPT-6.1 Terra Oct 26, 2026
Health
Hallucination
Safety
Alignment
- Respecting Auto-Review 0
- Respecting Warnings 29.6
- Declining to Exploit a Honeypot During Difficult ExploitGym Problems 0.07
- Blocker Deception 8.47
- Broken Search Tool 1.83
- Misaligned Behavior in Realistic Work Environments 2.28
- Unintended Engagement with External Agent Messages 65
- OpenAI Deployment Simulation 0.045
Monitorability
Bio and chem risk
Cyber risk
- ExploitBench 97.1
- ExploitBench - Internal Port 24.5
- SEC-bench Pro 83.3
- Sandbox Bench 22.7