Up to four models
GPT-5.6 Terra vs Claude Haiku 5.5
On the index (GPT-4 = 100), GPT-5.6 Terra scores 178 and Claude Haiku 5.5 174. Side by side on the same third-party tests, with price and context. The link in your address bar shares this exact comparison.
GPT-5.6 Terra
178162–195
WeightsClosed
ReleasedJul 2026
Price in / out$2.0 / $12
Context1.05M
Core tests taken6 of 9
Claude Haiku 5.5
174158–191
WeightsClosed
ReleasedOct 2026
Price in / out$0.10 / $0.50
Context1M
Core tests taken7 of 9
GPT-5.6 Terra leads Claude Haiku 5.5 by 4 points, ahead on 5 of 6 shared core tests, at 23× the price.
Test by testthird-party results
GPT-5.6 TerraClaude Haiku 5.5
Every result as a table
| Test | GPT-5.6 Terra | Claude Haiku 5.5 |
|---|---|---|
| Reasoning & knowledge | ||
| GPQA Diamond (Epoch AI) | 93.3% | 89.6% |
| LiveBench Data Analysis | 79.3% | 72.8% |
| LiveBench IF | 64.6% | 66.5% |
| LiveBench Language | 82.9% | 63.5% |
| LiveBench Reasoning | 90.6% | 84.6% |
| SimpleQA Verified | 43.2% | 23.8% |
| Maths | ||
| FrontierMath (tiers 1-3) | 86.0% | 75.1% |
| AIME-style maths (OTIS mock) | 99.7% | 98.9% |
| LiveBench Mathematics | 94.9% | 92.8% |
| ProofBench v1.1 | 74.0% | 89.0% |
| Chess puzzles | 47.0% | 55.0% |
| FrontierMath (tier 4) | 70.7% | 46.3% |
| Coding | ||
| Terminal-Bench 4.0 | 22.7% | 36.4% |
| LiveBench Agentic Coding | 54.9% | 57.2% |
| LiveBench Coding | 78.3% | 77.9% |
| Code Migration | 47.8% | 44.6% |
| CyberBench v1.1 | 72.1% | 75.5% |
| IOI | 87.6% | 47.3% |
| ProgramBench | 0.5% | 2.5% |
| Vibe Code Bench 1-100 | 14.8% | 20.4% |
| Vibe Code Bench v1.1 | 74.6% | 90.4% |
| Agents | ||
| APEX-Agents | 58.2% | 56.1% |
| SRE Bench | 9.5% | 5.7% |
| Terminal-Bench-Science | 10.0% | 10.0% |
| DTBench | 93.3% | 95.0% |
| Furniture assembly | 54.2% | 47.5% |
| LMCA | 54.9% | 51.1% |
| Mystery game puzzles | 35.0% | 30.0% |
| Professional work | ||
| Finance Agent (v2) | 54.4% | 54.1% |
| EMB | 66.2% | 59.0% |
| Legal Research Bench | 41.3% | 43.8% |
| MedCode | 43.4% | 51.5% |
| MedScribe | 82.9% | 89.5% |
| Public Benefits Bench v1.1 | 62.4% | 63.2% |
| SAGE | 47.0% | 45.8% |
| Tax Agent Bench | 30.0% | 31.9% |
| Novel problems | ||
| ARC-AGI-2 | 83.9% | 61.3% |
| ARC-AGI-1 | 96.5% | 89.0% |
| ARC-AGI-3 | 0.8% | 0.5% |
Tests at least two of these models have taken. Bold rows are the core tests.
