Up to four models
GPT-5.6 Terra vs Gemini 3.8 Flash
On the index (GPT-4 = 100), GPT-5.6 Terra scores 178 and Gemini 3.8 Flash 175. 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
Gemini 3.8 Flash175159–191
WeightsClosed
ReleasedSep 2026
Price in / out$0.82 / $4.13
Context1.05M
Core tests taken7 of 9
GPT-5.6 Terra and Gemini 3.8 Flash are too close to call (3 points apart, within the uncertainty), ahead on 2 of 6 shared core tests, at 2.7× the price.
Test by testthird-party results
GPT-5.6 TerraGemini 3.8 Flash
Gemini 3.8 FlashEvery result as a table
| Test | GPT-5.6 Terra | Gemini 3.8 Flash |
|---|---|---|
| Reasoning & knowledge | ||
| GPQA Diamond (Epoch AI) | 93.3% | 95.4% |
| GPQA Diamond (Vals AI) | 90.9% | 94.4% |
| MMLU Pro | 86.7% | 90.2% |
| MMMU Pro | 86.5% | 89.1% |
| SimpleQA Verified | 43.2% | 69.7% |
| Maths | ||
| FrontierMath (tiers 1-3) | 86.0% | 68.4% |
| AIME-style maths (OTIS mock) | 99.7% | 98.9% |
| ProofBench v1.1 | 74.0% | 48.0% |
| Chess puzzles | 47.0% | 58.0% |
| FrontierMath (tier 4) | 70.7% | 21.9% |
| Surface Evolver Bench | 83.8% | 76.9% |
| Coding | ||
| Terminal-Bench 4.0 | 22.7% | 19.7% |
| Code Migration | 47.8% | 36.5% |
| CyberBench v1.1 | 72.1% | 43.8% |
| IOI | 87.6% | 56.9% |
| LiveCodeBench | 85.9% | 89.5% |
| ProgramBench | 0.5% | 1.0% |
| SWE-bench | 95.4% | 80.0% |
| Vibe Code Bench 1-100 | 14.8% | 18.8% |
| Vibe Code Bench v1.1 | 74.6% | 78.6% |
| DeepSWE | 69.6% | 73.8% |
| WeirdML | 78.3% | 84.8% |
| Agents | ||
| APEX-Agents | 58.2% | 64.3% |
| SkillsBench | 58.9% | 58.0% |
| SRE Bench | 9.5% | 11.1% |
| Terminal-Bench-Science | 10.0% | 8.6% |
| DTBench | 93.3% | 95.7% |
| Furniture assembly | 54.2% | 31.7% |
| LMCA | 54.9% | 52.9% |
| Mystery game puzzles | 35.0% | 47.0% |
| Professional work | ||
| Finance Agent (v2) | 54.4% | 61.4% |
| EMB | 66.2% | 72.2% |
| LegalBench | 85.1% | 87.0% |
| Legal Research Bench | 41.3% | 38.9% |
| MedCode | 43.4% | 48.1% |
| MedScribe | 82.9% | 84.5% |
| MortgageTax | 67.3% | 65.3% |
| Public Benefits Bench v1.1 | 62.4% | 65.3% |
| SAGE | 47.0% | 35.1% |
| Tax Agent Bench | 30.0% | 32.1% |
| TaxEval v2 | 76.2% | 74.5% |
| Novel problems | ||
| ARC-AGI-2 | 83.9% | 89.2% |
| ARC-AGI-1 | 96.5% | 98.5% |
| ARC-AGI-3 | 0.8% | 10.4% |
Tests at least two of these models have taken. Bold rows are the core tests.
