Gemini 3.5 Flash vs Kimi K3
Gemini 3.5 Flash 
Kimi K3 Verdict
- Kimi K3 leads on quality: 66.5 vs 63.9.
- Gemini 3.5 Flash (medium) is stronger in agents & tools, instruction following, knowledge, multimodal, reasoning.
- Kimi K3 is stronger in coding, composite, human preference, long context, maths.
- Gemini 3.5 Flash (medium) is 1.8× cheaper ($3.38 vs $6.00 per 1M blended).
- Gemini 3.5 Flash (medium) streams 5.7× faster (209 vs 37 tokens per second).
| Metric | Gemini 3.5 Flash (medium) | Kimi K3 |
|---|---|---|
| BenchLeader Index | 63.9 | 66.5 |
| Agents & tools score | 67.8 | 67.5 |
| Coding score | 58.9 | 64.9 |
| Composite score | 71.4 | 74.1 |
| Human preference score | 67.4 | 71.2 |
| Instruction following score | 72.2 | – |
| Knowledge score | 73.9 | 68.1 |
| Long context score | 63.6 | 71.1 |
| Multimodal score | 67.6 | 65.1 |
| Reasoning score | 66.8 | – |
| Maths score | – | 78.1 |
| Blended price $/M | $3.38 | $6.00 |
| Output speed | 209 tok/s | 37 tok/s |
| Time to first answer | 17.0 s | 57.7 s |
| Context window | 1M | 1.0M |
| SciCode | – | 58.7% |
| APEX-Agents | – | 39.3% |
| FrontierCode | – | 44.2% |
| ProofBench | – | 87.0% |
| Epoch Capabilities Index | – | 157.6 |
| LMArena Text | 1476 | – |
| LMArena Hard Prompts | 1493 | – |
| LMArena Coding | 1503 | – |
| LMArena WebDev | 1491 | – |
| LMArena Vision | 1306 | – |
| 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 | 33.6 | 43.8 |
| IFBench | 74.6% | – |
| AA-LCR | 74.3% | 88.7% |
| MMMU-Pro | 83.9% | 80.5% |
| AA-Omniscience | 20.8 | 19.7 |
| Terminal-Bench Hard | 39.4% | – |
| GPQA Diamond (AA) | 92.1% | 93.5% |
| Humanity's Last Exam (AA) | 41.3% | 46.9% |
| SciCode (AA) | – | 59.5% |
| 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 |
| EQ-Bench 4 | – | 1339 |
Data as of 2026-09-09. Best configuration of each model; every score links to its source on the model pages.