Head-to-head · Updated Nov 3, 2026
Kimi vs Claude
The highest-scoring Kimi and Claude models, Kimi K3.1 Pro and Claude Opus 5.5, on independent tests for professional work, cost and speed.
Short answer
Claude Opus 5.5 leads on 10 of 12 quality scores, including the overall Intelligence Index.
- FinanceClaude Opus 5.559.4 vs 53.3
- Legal workClaude Opus 5.558.0 vs 49.2
- HealthcareClaude Opus 5.557.9 vs 51.6
- EngineeringClaude Opus 5.564.6 vs 50.4
- Strategy & operationsClaude Opus 5.561.4 vs 57.6
- Coding agentsKimi K3.1 Pro67.2% vs 59.6%
- Expert knowledgeClaude Opus 5.561.4% vs 52.5%
- Fewer made-up answersKimi K3.1 Pro37.3% vs 50.3%
- Lower costKimi K3.1 Pro$3,153 vs $8,910
- SpeedClaude Opus 5.596 tok/s vs 31 tok/s
Scores, cost and speed
| Measure | Kimi K3.1 Pro | Claude Opus 5.5 (Max, Default Fallback) |
|---|---|---|
| Overall | ||
| Intelligence IndexCombined score across nine evaluations | 46.0 | Leads: 58.4 |
| Professional work | ||
| Finance & accounting | 53.3 | Leads: 59.4 |
| Legal | 49.2 | Leads: 58.0 |
| Healthcare & medical | 51.6 | Leads: 57.9 |
| Engineering | 50.4 | Leads: 64.6 |
| Strategy & operations | 57.6 | Leads: 61.4 |
| Economics | 51.3 | Leads: 62.3 |
| Specific tests | ||
| Terminal-Bench 4.0Agentic coding tasks | Leads: 67.2% | 59.6% |
| FrontierCode 2.0Hard programming problems | 25.5% | Leads: 46.4% |
| Humanity’s Last ExamExpert-level questions | 52.5% | Leads: 61.4% |
| HealthBenchHealth conversations graded by physicians’ rubrics | — | 57.0% |
| GDPval-AAKnowledge-work deliverables, Elo | 1858 | Leads: 1866 |
| Hallucination rateAA-Omniscience; lower is better | Leads: 37.3% | 50.3% |
| Cost & speed | ||
| Input priceUS$ per 1M tokens | Leads: $2 | $4 |
| Output priceUS$ per 1M tokens | Leads: $10 | $20 |
| Cost to run the indexUS$, includes reasoning tokens | Leads: $3,153 | $8,910 |
| Output speedMedian tokens per second | 31 tok/s | Leads: 96 tok/s |
| Context windowTokens | 1M | 1M |
Scores, prices and speeds: Artificial Analysis, independently measured, checked Nov 3, 2026. Model names include the tested reasoning setting. Stubbi has not rerun these evaluations. Consumer apps such as ChatGPT, Claude and Gemini may use other models and add features like search and memory.
All measured Kimi and Claude models
| Model and setting | Released | Index | Price in / out |
|---|---|---|---|
| Claude Opus 5.5 (Max, Default Fallback)Anthropic | Sep 2026 | 58.4 | $4 / $20 |
| Claude Sonnet 5.5 (Max, Default Fallback)Anthropic | Sep 2026 | 54.2 | $2 / $10 |
| Claude Fable 5.1 (Max, Default Fallback)Anthropic | Sep 2026 | 49.0 | $10 / $50 |
| Kimi K3.1 ProKimi | Oct 2026 | 46.0 | $2 / $10 |
| Claude Haiku 5.5 (Max)Anthropic | Oct 2026 | 36.2 | $0.10 / $0.50 |
Test them on your own work
Public scores can’t show how each model handles your documents, clients or standards. Tell us the task, and we’ll consider a side-by-side comparison with the outputs you can judge.