Up to four models
Claude Opus 5 vs Kimi K3
On the index (GPT-4 = 100), Claude Opus 5 scores 185 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.
Claude Opus 5
185169–201
WeightsClosed
ReleasedJul 2026
Price in / out$5.0 / $25
Context1M
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
Claude Opus 5 leads Kimi K3 by 17 points, ahead on 6 of 6 shared core tests, at 2.4× the price.
Test by testthird-party results
Claude Opus 5Kimi K3
Every result as a table
| Test | Claude Opus 5 | Kimi K3 |
|---|---|---|
| Reasoning & knowledge | ||
| GPQA Diamond (Epoch AI) | 93.9% | 93.1% |
| LiveBench Data Analysis | 74.6% | 78.7% |
| LiveBench IF | 63.8% | 71.4% |
| LiveBench Language | 88.7% | 85.5% |
| LiveBench Reasoning | 91.2% | 90.7% |
| GPQA Diamond (Vals AI) | 93.4% | 92.9% |
| MMLU Pro | 91.6% | 88.0% |
| MMMU Pro | 89.9% | 88.1% |
| SimpleBench | 80.6% | 60.7% |
| SimpleQA Verified | 59.9% | 50.6% |
| Maths | ||
| FrontierMath (tiers 1-3) | 85.6% | 72.2% |
| AIME-style maths (OTIS mock) | 98.9% | 97.2% |
| LiveBench Mathematics | 95.7% | 84.4% |
| ProofBench v1.1 | 99.0% | 87.0% |
| Chess puzzles | 43.0% | 37.0% |
| FrontierMath (tier 4) | 73.2% | 39.0% |
| Coding | ||
| Terminal-Bench 4.0 | 53.5% | 17.2% |
| LiveBench Agentic Coding | 65.2% | 62.2% |
| LiveBench Coding | 81.5% | 81.5% |
| Code Migration | 57.5% | 16.1% |
| CyberBench v1.1 | 65.4% | 75.2% |
| IOI | 84.3% | 48.9% |
| LiveCodeBench | 89.0% | 87.2% |
| ProgramBench | 3.0% | 2.0% |
| SWE-bench | 97.0% | 93.4% |
| Vibe Code Bench 1-100 | 28.5% | 18.2% |
| Vibe Code Bench v1.1 | 88.4% | 85.0% |
| DeepSWE | 73.7% | 68.5% |
| FrontierSWE | 52.0% | 25.9% |
| MirrorCode | 87.8% | 54.4% |
| WeirdML | 91.8% | 82.6% |
| Agents | ||
| APEX-Agents | 65.8% | 47.8% |
| BioMysteryBench | 79.3% | 72.6% |
| Terminal-Bench-Science | 27.1% | 1.4% |
| DTBench | 97.9% | 91.2% |
| Furniture assembly | 60.8% | 34.2% |
| LMCA | 64.5% | 52.7% |
| Mystery game puzzles | 59.0% | 26.0% |
| Professional work | ||
| Finance Agent (v2) | 58.6% | 53.1% |
| CorpFin v2 | 73.2% | 71.6% |
| EMB | 73.6% | 66.7% |
| LegalBench | 87.0% | 86.2% |
| Legal Research Bench | 55.3% | 46.2% |
| MedCode | 63.6% | 49.4% |
| MedScribe | 91.0% | 88.0% |
| MortgageTax | 72.1% | 66.3% |
| Public Benefits Bench v1.1 | 76.9% | 68.2% |
| Tax Agent Bench | 47.1% | 33.4% |
| TaxEval v2 | 75.1% | 75.7% |
| Novel problems | ||
| ARC-AGI-2 | 90.4% | 60.4% |
| ARC-AGI-1 | 97.5% | 94.5% |
Tests at least two of these models have taken. Bold rows are the core tests.