Up to four models
Kimi K3 vs Claude Opus 4.7
On the index (GPT-4 = 100), Kimi K3 scores 168 and Claude Opus 4.7 167. Side by side on the same third-party tests, with price and context. The link in your address bar shares this exact comparison.
Kimi K3
168152–183
WeightsOpen
ReleasedJul 2026
Price in / out$0.19 / $16
Context1.05M
Core tests taken6 of 9
Claude Opus 4.7
167151–183
WeightsClosed
ReleasedApr 2026
Price in / out$5.0 / $25
Context1M
Core tests taken6 of 9
Kimi K3 and Claude Opus 4.7 are too close to call (1 point apart, within the uncertainty), ahead on 3 of 4 shared core tests, and costs 2.4× less.
Test by testthird-party results
Kimi K3Claude Opus 4.7
Every result as a table
| Test | Kimi K3 | Claude Opus 4.7 |
|---|---|---|
| Reasoning & knowledge | ||
| GPQA Diamond (Epoch AI) | 93.1% | 90.1% |
| LiveBench Data Analysis | 78.7% | 78.3% |
| LiveBench IF | 71.4% | 66.7% |
| LiveBench Language | 85.5% | 77.9% |
| LiveBench Reasoning | 90.7% | 87.2% |
| GPQA Diamond (Vals AI) | 92.9% | 90.1% |
| MMLU Pro | 88.0% | 89.9% |
| MMMU Pro | 88.1% | 85.5% |
| SimpleBench | 60.7% | 61.7% |
| SimpleQA Verified | 50.6% | 51.7% |
| Maths | ||
| FrontierMath (tiers 1-3) | 72.2% | 70.2% |
| AIME-style maths (OTIS mock) | 97.2% | 97.8% |
| LiveBench Mathematics | 84.4% | 92.8% |
| Chess puzzles | 37.0% | 41.0% |
| FrontierMath (tier 4) | 39.0% | 31.7% |
| Coding | ||
| LiveBench Agentic Coding | 62.2% | 50.7% |
| LiveBench Coding | 81.5% | 82.1% |
| Code Migration | 16.1% | 43.9% |
| LiveCodeBench | 87.2% | 85.1% |
| ProgramBench | 2.0% | 0.0% |
| SWE-bench | 93.4% | 82.0% |
| Vibe Code Bench v1.1 | 85.0% | 71.0% |
| MirrorCode | 54.4% | 43.3% |
| WeirdML | 82.6% | 76.4% |
| Agents | ||
| APEX-Agents | 47.8% | 49.2% |
| DTBench | 91.2% | 94.7% |
| Furniture assembly | 34.2% | 33.3% |
| LMCA | 52.7% | 52.2% |
| Mystery game puzzles | 26.0% | 28.0% |
| Professional work | ||
| Finance Agent (v2) | 53.1% | 51.5% |
| CorpFin v2 | 71.6% | 66.1% |
| EMB | 66.7% | 63.7% |
| LegalBench | 86.2% | 85.3% |
| Legal Research Bench | 46.2% | 38.5% |
| MedCode | 49.4% | 54.9% |
| MedScribe | 88.0% | 83.0% |
| MortgageTax | 66.3% | 70.3% |
| Tax Agent Bench | 33.4% | 23.7% |
| TaxEval v2 | 75.7% | 75.3% |
Tests at least two of these models have taken. Bold rows are the core tests.