Model comparison
Claude Haiku 4.5 vs Kimi K3
Kimi K3 is the stronger model overall, scoring 59.5 to 39.5 on the Noometry Index. Claude Haiku 4.5 costs 3.0× less per token, which makes it the better buy when Kimi K3's lead doesn't matter for your workload.
Last verified . 36 shared benchmarks.
Summary
- They share 36 benchmarks with published results for both. Claude Haiku 4.5 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 15.1.
- The biggest single-benchmark swing is NYT Connections (extended): 14.3% for Claude Haiku 4.5 and 93.6% for Kimi K3.
- Claude Haiku 4.5 is cheaper at $1 / $5 per million input/output tokens, against $3 / $15 for Kimi K3.
- 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 Haiku 4.5 | Kimi K3 | |
|---|---|---|
| Provider | Anthropic | Moonshot AI |
| Noometry Index | 39.5 | 59.5 |
| Released | 2025-10-15 | 2026-07-16 |
| Weights | Proprietary | Open |
| Context window | 200K | 1.05M |
| Max output | 64K | 1.05M |
| Input $ / M tokens | $1 | $3 |
| Output $ / M tokens | $5 | $15 |
| Results tracked | 53 | 53 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding Kimi K3 leads
Claude Haiku 4.5: 44.0 (#78), Kimi K3: 61.0 (#10)
| Benchmark | Claude Haiku 4.5 | Kimi K3 |
|---|---|---|
| LMArena WebDev | 1330 | 1654 |
| SciCode | 43.3% | 59.5% |
| WeirdML | 45.4% | 82.6% |
| LMArena Coding | 1453 | 1508 |
| ALE-Bench | 653.48 | 1,524 |
| DeepSWE | — | 68.5% |
| FrontierCode | — | 44.2% |
| SWE-bench Verified (bash only) | 66.6% | — |
| SWE-bench Multilingual | 64.7% | — |
| FrontierSWE | — | 25.9% |
Agentic & Tool Use Kimi K3 leads
Claude Haiku 4.5: 33.6 (#52), Kimi K3: 41.8 (#20)
| Benchmark | Claude Haiku 4.5 | Kimi K3 |
|---|---|---|
| Vending-Bench 2 | 458.89 | 5,165 |
| Terminal-Bench | 35.5% | — |
| APEX-Agents | — | 50.6% |
| Berkeley Function Calling Leaderboard | 68.7% | — |
| τ²-bench Banking | — | 37.1% |
| DeepResearch Bench | 45.5% | — |
| PostTrainBench | — | 32% |
| BALROG | 31.2% | — |
| ExploitBench | 13.7% | — |
| GBAEval | — | 48.3% |
| GDP.pdf | — | 19% |
Reasoning Kimi K3 leads
Claude Haiku 4.5: 15.1 (#320), Kimi K3: 63.0 (#17)
| Benchmark | Claude Haiku 4.5 | Kimi K3 |
|---|---|---|
| ARC-AGI-2 | 4% | 60.4% |
| NYT Connections (extended) | 14.3% | 93.6% |
| ARC-AGI-1 | 47.7% | 94.5% |
| CritPt | 0% | 23.4% |
| Chess Puzzles | 8% | 39% |
| LMArena Hard Prompts | 1420 | 1496 |
| DTBench | 73.6% | 91.2% |
| LMCA | 30.9% | 52.7% |
| Epoch Capabilities Index | 142.41 | 157.45 |
| ForecastBench | 61.4 | 61.1 |
| SimpleBench | — | 60.7% |
| Mystery Game Puzzles | — | 26% |
| Surface Evolver Bench | — | 95% |
Math Kimi K3 leads
Claude Haiku 4.5: 44.9 (#78), Kimi K3: 74.2 (#16)
| Benchmark | Claude Haiku 4.5 | Kimi K3 |
|---|---|---|
| OTIS Mock AIME 2024-2025 | 66.7% | 97.2% |
| LMArena Math | 1396 | 1491 |
| FrontierMath (Tiers 1-3) | — | 72.2% |
| FrontierMath Tier 4 | — | 39% |
| MathArena Final-Answer Competitions | — | 87.8% |
| ProofBench | — | 87% |
| Omni-MATH | 56.1% | — |
| MATH Level 5 | 96.4% | — |
| FrontierMath (Feb 2025 set) | 5.9% | — |
| FrontierMath Tier 4 (v1) | 2.1% | — |
Knowledge Kimi K3 leads
Claude Haiku 4.5: 37.7 (#153), Kimi K3: 63.2 (#21)
| Benchmark | Claude Haiku 4.5 | Kimi K3 |
|---|---|---|
| GPQA Diamond | 71.2% | 93.1% |
| SimpleQA Verified | 13.2% | 50.6% |
| LMArena Expert | 1442 | 1521 |
| MMLU-Pro | 77.7% | — |
| Vectara Hallucination Rate | 9.8% | — |
| GPQA (HELM) | 60.5% | — |
Multimodal Kimi K3 leads
Claude Haiku 4.5: 26.8 (#118), Kimi K3: 37.8 (#70)
| Benchmark | Claude Haiku 4.5 | Kimi K3 |
|---|---|---|
| Blueprint-Bench 2 | 0% | 29.5% |
| Furniture Assembly | — | 34.2% |
| LMArena Document | 1420 | — |
Multilingual Kimi K3 leads
Claude Haiku 4.5: 49.9 (#129), Kimi K3: 56.3 (#21)
| Benchmark | Claude Haiku 4.5 | Kimi K3 |
|---|---|---|
| LMArena Non-English | 1377 | 1466 |
| LMArena Chinese | 1417 | 1529 |
| LMArena French | 1408 | 1491 |
| LMArena German | 1375 | 1488 |
| LMArena Japanese | 1339 | 1487 |
| LMArena Korean | 1347 | 1458 |
| LMArena Russian | 1381 | 1482 |
| LMArena Spanish | 1420 | 1472 |
Instruction Following Kimi K3 leads
Claude Haiku 4.5: 71.4 (#149), Kimi K3: 77.7 (#14)
| Benchmark | Claude Haiku 4.5 | Kimi K3 |
|---|---|---|
| LMArena Instruction Following | 1414 | 1483 |
| IFEval | 80.1% | — |
Long Context Kimi K3 leads
Claude Haiku 4.5: 43.6 (#92), Kimi K3: 45.8 (#29)
| Benchmark | Claude Haiku 4.5 | Kimi K3 |
|---|---|---|
| LMArena Longer Query | 1427 | 1494 |
Writing & Preference Kimi K3 leads
Claude Haiku 4.5: 57.9 (#123), Kimi K3: 76.6 (#4)
| Benchmark | Claude Haiku 4.5 | Kimi K3 |
|---|---|---|
| LMArena Text | 1396 | 1476 |
| LMArena Creative Writing | 1372 | 1454 |
| EQ-Bench 4 | 1064 | 1339 |
| LMArena Multi-Turn | 1409 | 1488 |
| EQ-Bench Creative Writing | — | 2082 |
| WildBench | 83.9% | — |
Frequently asked questions
Is Claude Haiku 4.5 better than Kimi K3?
Kimi K3 is the stronger model overall, scoring 59.5 to 39.5 on the Noometry Index. Claude Haiku 4.5 costs 3.0× less per token, which makes it the better buy when Kimi K3's lead doesn't matter for your workload.
Which is cheaper, Claude Haiku 4.5 or Kimi K3?
Claude Haiku 4.5 is cheaper. It lists at $1 per million input tokens and $5 per million output tokens; Kimi K3 lists at $3 and $15.
Is Claude Haiku 4.5 or Kimi K3 better for coding?
Kimi K3 scores higher on coding benchmarks: 61.0 versus 44.0 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 Haiku 4.5 and Kimi K3 share?
36 benchmarks have published results for both models. Claude Haiku 4.5 has 53 scored results on Noometry and Kimi K3 has 53.