Kimi K3 vs Qwen3.5 397B A17B
Kimi K3 
Qwen3.5 397B A17B Verdict
- Kimi K3 leads on quality: 66.5 vs 58.7.
- Kimi K3 is stronger in coding, composite, human preference, knowledge, long context, maths, multimodal.
- Qwen3.5 397B A17B is stronger in agents & tools, instruction following, reasoning.
- Qwen3.5 397B A17B is 16× cheaper ($0.387 vs $6.00 per 1M blended).
- Qwen3.5 397B A17B streams 2.1× faster (78 vs 38 tokens per second).
| Metric | Kimi K3 | Qwen3.5 397B A17B |
|---|---|---|
| BenchLeader Index | 66.5 | 58.7 |
| Agents & tools score | 67.4 | 69.4 |
| Coding score | 65.0 | 51.7 |
| Composite score | 74.2 | 53.1 |
| Human preference score | 71.2 | 63.6 |
| Knowledge score | 68.2 | 49.5 |
| Long context score | 71.1 | 65.2 |
| Maths score | 78.1 | 50.1 |
| Multimodal score | 65.0 | 61.6 |
| Instruction following score | – | 76.1 |
| Reasoning score | – | 63.0 |
| Blended price $/M | $6.00 | $0.387 |
| Output speed | 38 tok/s | 78 tok/s |
| Time to first answer | 56.3 s | 42.7 s |
| Context window | 1.0M | 262k |
| GPQA Diamond | – | 85.9% |
| FrontierMath Tiers 1–3 | – | 29.5% |
| OTIS Mock AIME | – | 88.9% |
| SciCode | 58.7% | – |
| APEX-Agents | 39.3% | – |
| FrontierCode | 44.2% | – |
| ProofBench | 87.0% | – |
| Epoch Capabilities Index | 157.6 | 147.0 |
| LMArena Text | – | 1441 |
| LMArena Hard Prompts | – | 1463 |
| LMArena Coding | – | 1491 |
| LMArena WebDev | – | 1399 |
| LMArena Vision | – | 1265 |
| LiveBench | 79.2% | – |
| LiveBench Reasoning | 90.7% | – |
| LiveBench Coding | 81.5% | – |
| LiveBench Agentic Coding | 62.2% | – |
| LiveBench Mathematics | 84.4% | – |
| LiveBench Data Analysis | 78.7% | – |
| LiveBench Language | 85.5% | – |
| AA Intelligence Index | 43.8 | 19.1 |
| IFBench | – | 78.8% |
| AA-LCR | 88.7% | 77.3% |
| MMMU-Pro | 80.5% | 77.3% |
| AA-Omniscience | 19.7 | -30.8 |
| Terminal-Bench Hard | – | 40.9% |
| GPQA Diamond (AA) | 93.5% | 89.3% |
| Humanity's Last Exam (AA) | 46.9% | 29.0% |
| SciCode (AA) | 59.5% | 44.8% |
| τ²-Bench Telecom (AA) | – | 95.6% |
| LegalBench | 86.0% | – |
| CorpFin | 71.6% | – |
| TaxEval | 75.7% | – |
| Terminal-Bench 2.1 (Vals) | 80.9% | – |
| SWE-bench (Vals) | 93.4% | – |
| GPQA Diamond (Vals) | 92.9% | – |
| Vals Index | 57.8 | – |
| AIME 2026 | – | 94.2% |
| HMMT February 2026 | – | 87.9% |
| EQ-Bench 4 | 1339 | – |
Data as of 2026-09-10. Best configuration of each model; every score links to its source on the model pages.
Kimi K3 vs Qwen3.5 397B A17B: questions
- Is Kimi K3 better than Qwen3.5 397B A17B?
- Kimi K3 leads on quality: 66.5 vs 58.7. The BenchLeader Index combines every independent quality benchmark; Kimi K3 is ahead overall as of 2026-09-10, but check the category scores for your use.
- Is Kimi K3 better than Qwen3.5 397B A17B for coding?
- Kimi K3 scores higher in coding (65 vs 52 on the category index, where 50 is average).
- Is Kimi K3 better than Qwen3.5 397B A17B for agentic tasks?
- Qwen3.5 397B A17B scores higher in agentic tasks (69 vs 67 on the category index, where 50 is average).
- Which is cheaper, Kimi K3 or Qwen3.5 397B A17B?
- Qwen3.5 397B A17B is cheaper: $0.387 against $6.00 per million tokens, blended at three input tokens per output token.
- Which is faster, Kimi K3 or Qwen3.5 397B A17B?
- Qwen3.5 397B A17B streams faster: 78 against 38 output tokens per second.
- Which has the larger context window?
- Kimi K3 accepts more context: 1.0M against 262k tokens.