Up to four models
GPT-5.6 Sol vs Kimi K3
On the index (GPT-4 = 100), GPT-5.6 Sol scores 179 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 Sol
179164–195
WeightsClosed
ReleasedJul 2026
Price in / out$4.0 / $20
Context1.05M
Core tests taken5 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 Sol leads Kimi K3 by 12 points, ahead on 5 of 5 shared core tests, at 1.9× the price.
Test by testthird-party results
GPT-5.6 SolKimi K3
Every result as a table
| Test | GPT-5.6 Sol | Kimi K3 |
|---|---|---|
| Reasoning & knowledge | ||
| GPQA Diamond (Epoch AI) | 93.5% | 93.1% |
| LiveBench Data Analysis | 79.8% | 78.7% |
| LiveBench IF | 71.9% | 71.4% |
| LiveBench Language | 87.7% | 85.5% |
| LiveBench Reasoning | 91.6% | 90.7% |
| GPQA Diamond (Vals AI) | 95.2% | 92.9% |
| MMLU Pro | 89.1% | 88.0% |
| MMMU Pro | 88.8% | 88.1% |
| SimpleBench | 64.8% | 60.7% |
| SimpleQA Verified | 69.7% | 50.6% |
| Maths | ||
| FrontierMath (tiers 1-3) | 89.1% | 72.2% |
| AIME-style maths (OTIS mock) | 100.0% | 97.2% |
| LiveBench Mathematics | 96.2% | 84.4% |
| ProofBench v1.1 | 83.0% | 87.0% |
| Chess puzzles | 59.0% | 37.0% |
| FrontierMath (tier 4) | 82.9% | 39.0% |
| Surface Evolver Bench | 93.1% | 95.0% |
| Coding | ||
| Terminal-Bench 4.0 | 37.9% | 17.2% |
| LiveBench Agentic Coding | 56.2% | 62.2% |
| LiveBench Coding | 83.9% | 81.5% |
| Code Migration | 52.9% | 16.1% |
| CyberBench v1.1 | 76.3% | 75.2% |
| IOI | 91.2% | 48.9% |
| LiveCodeBench | 82.6% | 87.2% |
| ProgramBench | 1.5% | 2.0% |
| SWE-bench | 96.2% | 93.4% |
| Vibe Code Bench 1-100 | 20.0% | 18.2% |
| Vibe Code Bench v1.1 | 80.5% | 85.0% |
| DeepSWE | 72.7% | 68.5% |
| FrontierSWE | 32.2% | 25.9% |
| MirrorCode | 20.0% | 54.4% |
| WeirdML | 88.8% | 82.6% |
| Agents | ||
| BioMysteryBench | 71.1% | 72.6% |
| Terminal-Bench-Science | 20.0% | 1.4% |
| DTBench | 95.7% | 91.2% |
| Furniture assembly | 56.7% | 34.2% |
| LMCA | 58.5% | 52.7% |
| Mystery game puzzles | 58.0% | 26.0% |
| Professional work | ||
| Finance Agent (v2) | 53.8% | 53.1% |
| CorpFin v2 | 64.4% | 71.6% |
| EMB | 72.3% | 66.7% |
| LegalBench | 87.0% | 86.2% |
| Legal Research Bench | 48.1% | 46.2% |
| MedCode | 44.0% | 49.4% |
| MedScribe | 85.2% | 88.0% |
| MortgageTax | 67.3% | 66.3% |
| Public Benefits Bench v1.1 | 66.5% | 68.2% |
| Tax Agent Bench | 31.1% | 33.4% |
| TaxEval v2 | 74.8% | 75.7% |
| Novel problems | ||
| ARC-AGI-2 | 92.5% | 60.4% |
| ARC-AGI-1 | 97.5% | 94.5% |
Tests at least two of these models have taken. Bold rows are the core tests.