Grok 4.20 vs Kimi K3
Grok 4.20 
Kimi K3 Verdict
- Kimi K3 (max) leads on quality: 67.2 vs 59.9.
- Grok 4.20 (thinking) is stronger in instruction following.
- Kimi K3 (max) is stronger in agents & tools, coding, composite, human preference, knowledge, long context, maths, multimodal, reasoning.
- Grok 4.20 (thinking) is 3.8× cheaper ($1.56 vs $6.00 per 1M blended).
- Grok 4.20 (thinking) streams 2.7× faster (101 vs 37 tokens per second).
| Metric | Grok 4.20 (thinking) | Kimi K3 (max) |
|---|---|---|
| BenchLeader Index | 59.9 | 67.2 |
| Agents & tools score | 51.7 | 67.5 |
| Coding score | 54.7 | 67.0 |
| Composite score | 61.5 | 84.5 |
| Human preference score | 67.1 | 68.8 |
| Instruction following score | 79.9 | – |
| Knowledge score | 56.8 | 64.5 |
| Long context score | 60.7 | 71.0 |
| Maths score | 55.6 | 65.4 |
| Multimodal score | 59.9 | 65.1 |
| Reasoning score | 60.7 | 68.6 |
| Blended price $/M | $1.56 | $6.00 |
| Output speed | 101 tok/s | 37 tok/s |
| Time to first answer | 21.4 s | 58.3 s |
| Context window | 1M | 1.0M |
| GPQA Diamond | 89.3% | 93.1% |
| FrontierMath Tiers 1–3 | 44.9% | 72.2% |
| FrontierMath Tier 4 | 17.1% | 39.0% |
| OTIS Mock AIME | 92.2% | 97.2% |
| SimpleQA Verified | 30.2% | 50.6% |
| SimpleBench | – | 60.7% |
| SciCode | – | 58.7% |
| WeirdML | – | 82.6% |
| ProofBench | 14.0% | – |
| LMArena Text | 1472 | 1485 |
| LMArena Hard Prompts | 1488 | 1514 |
| LMArena Coding | 1510 | 1538 |
| LMArena WebDev | 1374 | 1674 |
| LMArena Vision | 1263 | – |
| LMArena Agent | – | 6.2 |
| AA Intelligence Index | 25.7 | 43.8 |
| IFBench | 82.9% | – |
| AA-LCR | 69.0% | 88.7% |
| MMMU-Pro | 74.6% | 80.5% |
| AA-Omniscience | 14.8 | 19.7 |
| Terminal-Bench Hard | 40.9% | – |
| GPQA Diamond (AA) | 91.1% | 93.5% |
| Humanity's Last Exam (AA) | 34.5% | 46.9% |
| SciCode (AA) | – | 59.5% |
| τ²-Bench Telecom (AA) | 96.5% | – |
| AIME (Vals) | 96.5% | – |
| LiveCodeBench | 84.3% | 87.2% |
| MMLU-Pro | 86.3% | 88.0% |
| IOI | – | 48.9% |
| LegalBench | 77.7% | – |
| CorpFin | 63.7% | – |
| TaxEval | 74.1% | – |
| MedQA | 94.5% | – |
| Terminal-Bench 2.1 (Vals) | 44.2% | – |
| SWE-bench (Vals) | 72.2% | – |
| GPQA Diamond (Vals) | 88.6% | – |
| Vals Index | 17.6 | – |
| MCP Atlas | – | 82.3% |
| Kagi LLM Benchmark | 75.0% | – |
| ARC-AGI-1 | 89.5% | 94.5% |
| ARC-AGI-2 | 65.1% | 60.4% |
| ARC-AGI-3 | 0.1% | – |
| CritPt | 6.6% | 23.4% |
| GDPval (AA) | – | 52.4% |
| τ²-Bench Banking (AA) | – | 46.0% |
| ITBench SRE (AA) | – | 47.7% |
| Analyst Agent (AA) | – | 38.8% |
| APEX-Agents (AA) | 14.2% | 41.3% |
| BioMysteryBench | – | 71.5% |
| CaseLaw v2 | 54.5% | – |
| Code Migration | 0.3% | 16.1% |
| Excel Modeling Benchmark | 11.7% | 66.4% |
| Finance Agent v2 | 28.5% | – |
| Harvey's Legal Agent Benchmark | 0.0% | – |
| Legal Research Bench | 13.9% | – |
| MedCode | 32.2% | – |
| MedScribe | 63.4% | – |
| MMMU-Pro (Vals) | 83.5% | – |
| MortgageTax | 45.4% | – |
| SAGE | 38.2% | – |
| Tax Agent Bench | – | 68.7% |
| Terminal-Bench 2.0 (Vals) | 40.5% | – |
| Terminal-Bench 4.0 (Vals) | – | 12.6% |
| Terminal-Bench Science | – | 2.9% |
| Time Horizon Index: KSP | – | 10.5% |
| Vals Multimodal Index | 39.1% | – |
| Vibe Code Bench 1-100 | – | 18.2% |
| Vibe Code Bench v1.1 | 4.1% | – |
| FORTRESS | – | 26.6% |
| LMArena Maths | 1467 | 1501 |
| LMArena Creative Writing | 1446 | 1458 |
| LMArena Instruction Following | 1445 | 1484 |
| LMArena Multi-turn | 1480 | 1496 |
| LMArena Longer Queries | 1465 | 1500 |
| LMArena Document | 1439 | – |
| Chess Puzzles | 24.0% | 39.0% |
| Mystery Game Puzzles | – | 26.0% |
| Surface Evolver Bench | – | 93.0% |
| DeepSWE | – | 68.5% |
| LMCA | – | 52.7% |
| DTBench | – | 91.2% |
| ForecastBench | 60.7% | – |
| CursorBench | – | 60.8% |
| ALE-Bench | – | 1524.5 |
| GDP.pdf | – | 19.0% |
| FrontierSWE | – | 25.9% |
Data as of 2026-09-19. Best configuration of each model; every score links to its source on the model pages.
Grok 4.20 vs Kimi K3: questions
- Is Grok 4.20 better than Kimi K3?
- Kimi K3 (max) leads on quality: 67.2 vs 59.9. The BenchLeader Index combines every independent quality benchmark; Kimi K3 (max) is ahead overall as of 2026-09-19, but check the category scores for your use.
- Is Grok 4.20 better than Kimi K3 for coding?
- Kimi K3 scores higher in coding (67 vs 55 on the category index, where 50 is average).
- Is Grok 4.20 better than Kimi K3 for agentic tasks?
- Kimi K3 scores higher in agentic tasks (68 vs 52 on the category index, where 50 is average).
- Which is cheaper, Grok 4.20 or Kimi K3?
- Grok 4.20 is cheaper: $1.56 against $6.00 per million tokens, blended at three input tokens per output token.
- Which is faster, Grok 4.20 or Kimi K3?
- Grok 4.20 streams faster: 101 against 37 output tokens per second.
- Which has the larger context window?
- Kimi K3 accepts more context: 1.0M against 1M tokens.