Up to four models
GPT-5.6 Terra vs Kimi K3
On the index (GPT-4 = 100), GPT-5.6 Terra scores 172 and Kimi K3 168. 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
172156–188
WeightsClosed
ReleasedJul 2026
Price in / out$2.0 / $12
Context1.05M
Core tests taken6 of 9
Kimi K3
168152–183
WeightsOpen
ReleasedJul 2026
Price in / out$0.19 / $16
Context1.05M
Core tests taken6 of 9
GPT-5.6 Terra and Kimi K3 are too close to call (4 points apart, within the uncertainty), ahead on 6 of 6 shared core tests, at a similar price.
Test by testthird-party results
GPT-5.6 TerraKimi K3
Every result as a table
| Test | GPT-5.6 Terra | Kimi K3 |
|---|---|---|
| Reasoning & knowledge | ||
| GPQA Diamond (Epoch AI) | 93.3% | 93.1% |
| LiveBench Data Analysis | 79.3% | 78.7% |
| LiveBench IF | 64.6% | 71.4% |
| LiveBench Language | 82.9% | 85.5% |
| LiveBench Reasoning | 90.6% | 90.7% |
| GPQA Diamond (Vals AI) | 90.9% | 92.9% |
| MMLU Pro | 86.7% | 88.0% |
| MMMU Pro | 86.5% | 88.1% |
| SimpleBench | 48.9% | 60.7% |
| SimpleQA Verified | 43.2% | 50.6% |
| Maths | ||
| FrontierMath (tiers 1-3) | 86.0% | 72.2% |
| AIME-style maths (OTIS mock) | 99.7% | 97.2% |
| LiveBench Mathematics | 94.9% | 84.4% |
| ProofBench v1.1 | 74.0% | 87.0% |
| Chess puzzles | 47.0% | 37.0% |
| FrontierMath (tier 4) | 70.7% | 39.0% |
| Surface Evolver Bench | 83.8% | 95.0% |
| Coding | ||
| Terminal-Bench 4.0 | 22.7% | 17.2% |
| LiveBench Agentic Coding | 54.9% | 62.2% |
| LiveBench Coding | 78.3% | 81.5% |
| Code Migration | 47.8% | 16.1% |
| CyberBench v1.1 | 72.1% | 75.2% |
| IOI | 87.6% | 48.9% |
| LiveCodeBench | 85.9% | 87.2% |
| ProgramBench | 0.5% | 2.0% |
| SWE-bench | 95.4% | 93.4% |
| Vibe Code Bench 1-100 | 14.8% | 18.2% |
| Vibe Code Bench v1.1 | 74.6% | 85.0% |
| DeepSWE | 69.6% | 68.5% |
| MirrorCode | 41.1% | 54.4% |
| WeirdML | 78.3% | 82.6% |
| Agents | ||
| APEX-Agents | 58.2% | 47.8% |
| Terminal-Bench-Science | 10.0% | 1.4% |
| DTBench | 93.3% | 91.2% |
| Furniture assembly | 54.2% | 34.2% |
| LMCA | 54.9% | 52.7% |
| Mystery game puzzles | 35.0% | 26.0% |
| Professional work | ||
| Finance Agent (v2) | 54.4% | 53.1% |
| CorpFin v2 | 65.3% | 71.6% |
| EMB | 66.2% | 66.7% |
| LegalBench | 85.1% | 86.2% |
| Legal Research Bench | 41.3% | 46.2% |
| MedCode | 43.4% | 49.4% |
| MedScribe | 82.9% | 88.0% |
| MortgageTax | 67.3% | 66.3% |
| Public Benefits Bench v1.1 | 62.4% | 68.2% |
| SAGE | 87.0% | 86.8% |
| Tax Agent Bench | 30.0% | 33.4% |
| TaxEval v2 | 76.2% | 75.7% |
| Novel problems | ||
| ARC-AGI-2 | 83.9% | 60.4% |
| ARC-AGI-1 | 96.5% | 94.5% |
Tests at least two of these models have taken. Bold rows are the core tests.