Model comparison
Claude Sonnet 4 vs Kimi K3
Kimi K3 is the stronger model overall, scoring 59.5 to 40.8 on the Noometry Index.
Last verified . 31 shared benchmarks.
Summary
- They share 31 benchmarks with published results for both. Claude Sonnet 4 scores higher in 0 categories and Kimi K3 in 10 categories; 10 gaps are clear of the uncertainty.
- The widest gap is in reasoning, where Kimi K3 leads 63.0 to 22.9.
- The biggest single-benchmark swing is ARC-AGI-1: 40% for Claude Sonnet 4 and 94.5% for Kimi K3.
- Both cost about the same: $3 input and $15 output per million tokens.
- Kimi K3 accepts more context: 1.05M tokens versus 200K.
- Kimi K3 has downloadable open weights; the other is API-only.
Side by side
| Claude Sonnet 4 | Kimi K3 | |
|---|---|---|
| Provider | Anthropic | Moonshot AI |
| Noometry Index | 40.8 | 59.5 |
| Released | 2025-05-22 | 2026-07-16 |
| Weights | Proprietary | Open |
| Context window | 200K | 1.05M |
| Max output | 64K | 1.05M |
| Input $ / M tokens | $3 | $3 |
| Output $ / M tokens | $15 | $15 |
| Results tracked | 58 | 53 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding Kimi K3 leads
Claude Sonnet 4: 43.5 (#88), Kimi K3: 61.0 (#10)
| Benchmark | Claude Sonnet 4 | Kimi K3 |
|---|---|---|
| SciCode | 40% | 59.5% |
| WeirdML | 46.1% | 82.6% |
| LMArena Coding | 1414 | 1508 |
| ALE-Bench | 655.35 | 1,524 |
| DeepSWE | — | 68.5% |
| FrontierCode | — | 44.2% |
| SWE-bench Verified (bash only) | 64.9% | — |
| Aider Polyglot | 61.3% | — |
| LMArena WebDev | — | 1654 |
| FrontierSWE | — | 25.9% |
| GSO | 4.9% | — |
Agentic & Tool Use Kimi K3 leads
Claude Sonnet 4: 38.5 (#31), Kimi K3: 41.8 (#20)
| Benchmark | Claude Sonnet 4 | Kimi K3 |
|---|---|---|
| APEX-Agents | — | 50.6% |
| TheAgentCompany | 33.1% | — |
| τ²-bench Banking | — | 37.1% |
| Cybench | 35% | — |
| DeepResearch Bench | 46.6% | — |
| OSWorld | 43.9% | — |
| PostTrainBench | — | 32% |
| GBAEval | — | 48.3% |
| GDP.pdf | — | 19% |
| METR Time Horizons | 62% | — |
| Vending-Bench 2 | — | 5,165 |
Reasoning Kimi K3 leads
Claude Sonnet 4: 22.9 (#187), Kimi K3: 63.0 (#17)
| Benchmark | Claude Sonnet 4 | Kimi K3 |
|---|---|---|
| ARC-AGI-2 | 5.9% | 60.4% |
| SimpleBench | 45.5% | 60.7% |
| ARC-AGI-1 | 40% | 94.5% |
| CritPt | 0.3% | 23.4% |
| LMArena Hard Prompts | 1372 | 1496 |
| DTBench | 77.1% | 91.2% |
| LMCA | 29% | 52.7% |
| Epoch Capabilities Index | 141.69 | 157.45 |
| ForecastBench | 60.2 | 61.1 |
| Kagi LLM Benchmark | 73% | — |
| NYT Connections (extended) | — | 93.6% |
| Chess Puzzles | — | 39% |
| EnigmaEval | 3.1% | — |
| Mystery Game Puzzles | — | 26% |
| Surface Evolver Bench | — | 95% |
Math Kimi K3 leads
Claude Sonnet 4: 43.3 (#80), Kimi K3: 74.2 (#16)
| Benchmark | Claude Sonnet 4 | Kimi K3 |
|---|---|---|
| OTIS Mock AIME 2024-2025 | 71.1% | 97.2% |
| LMArena Math | 1375 | 1491 |
| FrontierMath (Tiers 1-3) | — | 72.2% |
| FrontierMath Tier 4 | — | 39% |
| MathArena Final-Answer Competitions | — | 87.8% |
| ProofBench | — | 87% |
| Omni-MATH | 60.2% | — |
| MATH Level 5 | 84.4% | — |
| FrontierMath (Feb 2025 set) | 4.1% | — |
| FrontierMath Tier 4 (v1) | 0% | — |
Knowledge Kimi K3 leads
Claude Sonnet 4: 41.8 (#108), Kimi K3: 63.2 (#21)
| Benchmark | Claude Sonnet 4 | Kimi K3 |
|---|---|---|
| GPQA Diamond | 79.2% | 93.1% |
| LMArena Expert | 1372 | 1521 |
| Humanity's Last Exam | 7.8% | — |
| SimpleQA Verified | — | 50.6% |
| MMLU-Pro | 84.3% | — |
| Confabulations | 13.2% | — |
| Vectara Hallucination Rate | 10.3% | — |
| GPQA (HELM) | 70.6% | — |
Multimodal Kimi K3 leads
Claude Sonnet 4: 26.2 (#121), Kimi K3: 37.8 (#70)
| Benchmark | Claude Sonnet 4 | Kimi K3 |
|---|---|---|
| LMArena Vision | 1191 | — |
| GeoBench | 37% | — |
| VPCT | 34% | — |
| Blueprint-Bench 2 | — | 29.5% |
| Furniture Assembly | — | 34.2% |
| MindCube | 44.8% | — |
Multilingual Kimi K3 leads
Claude Sonnet 4: 46.7 (#156), Kimi K3: 56.3 (#21)
| Benchmark | Claude Sonnet 4 | Kimi K3 |
|---|---|---|
| LMArena Non-English | 1333 | 1466 |
| LMArena Chinese | 1350 | 1529 |
| LMArena French | 1363 | 1491 |
| LMArena German | 1331 | 1488 |
| LMArena Japanese | 1302 | 1487 |
| LMArena Korean | 1291 | 1458 |
| LMArena Russian | 1355 | 1482 |
| LMArena Spanish | 1357 | 1472 |
Instruction Following Kimi K3 leads
Claude Sonnet 4: 71.7 (#145), Kimi K3: 77.7 (#14)
| Benchmark | Claude Sonnet 4 | Kimi K3 |
|---|---|---|
| LMArena Instruction Following | 1376 | 1483 |
| IFEval | 84% | — |
Long Context Kimi K3 leads
Claude Sonnet 4: 33.7 (#259), Kimi K3: 45.8 (#29)
| Benchmark | Claude Sonnet 4 | Kimi K3 |
|---|---|---|
| LMArena Longer Query | 1398 | 1494 |
| Fiction.LiveBench | 46.9% | — |
Writing & Preference Kimi K3 leads
Claude Sonnet 4: 57.1 (#132), Kimi K3: 76.6 (#4)
| Benchmark | Claude Sonnet 4 | Kimi K3 |
|---|---|---|
| LMArena Text | 1351 | 1476 |
| LMArena Creative Writing | 1345 | 1454 |
| EQ-Bench Creative Writing | 1483 | 2082 |
| LMArena Multi-Turn | 1376 | 1488 |
| Short-Story Creative Writing | 81.4% | — |
| WildBench | 83.8% | — |
| EQ-Bench 4 | — | 1339 |
Frequently asked questions
Is Claude Sonnet 4 better than Kimi K3?
Kimi K3 is the stronger model overall, scoring 59.5 to 40.8 on the Noometry Index.
Which is cheaper, Claude Sonnet 4 or Kimi K3?
Kimi K3 is cheaper. It lists at $3 per million input tokens and $15 per million output tokens; Claude Sonnet 4 lists at $3 and $15.
Is Claude Sonnet 4 or Kimi K3 better for coding?
Kimi K3 scores higher on coding benchmarks: 61.0 versus 43.5 in the Noometry coding category.
Which has the bigger context window?
Kimi K3 does, with 1.05M tokens against 200K.
How many benchmarks do Claude Sonnet 4 and Kimi K3 share?
31 benchmarks have published results for both models. Claude Sonnet 4 has 58 scored results on Noometry and Kimi K3 has 53.