# Claude Haiku 4.5 vs Kimi K2.5

> Kimi K2.5 is the stronger model overall, scoring 48.1 to 39.5 on the Noometry Index.

- Canonical page: https://noometry.com/compare/claude-haiku-4-5-vs-kimi-k2-5
- Last updated: 2026-10-12
- Shared benchmarks: 38

## Summary

- They share 38 benchmarks with published results for both. Claude Haiku 4.5 scores higher in 0 categories and Kimi K2.5 in 10 categories; 9 gaps are clear of the uncertainty.
- The widest gap is in reasoning, where Kimi K2.5 leads 31.2 to 15.1.
- The biggest single-benchmark swing is NYT Connections (extended): 14.3% for Claude Haiku 4.5 and 69.9% for Kimi K2.5.
- Kimi K2.5 is cheaper at $0.45 / $2.25 per million input/output tokens, against $1 / $5 for Claude Haiku 4.5.
- Kimi K2.5 accepts more context: 262K tokens versus 200K.
- Kimi K2.5 has downloadable open weights; the other is API-only.

## Snapshot

| | Claude Haiku 4.5 | Kimi K2.5 |
|---|---|---|
| Provider | Anthropic | Moonshot AI |
| Noometry Index | 39.5 | 48.1 |
| Rank | 165 | 57 |
| Context | 200K | 262K |
| Input $/M | $1 | $0.45 |
| Output $/M | $5 | $2.25 |
| Weights | Proprietary | Open |

## Coding

- Claude Haiku 4.5: 44.0 (#78)
- Kimi K2.5: 48.8 (#53)

| Benchmark | Claude Haiku 4.5 | Kimi K2.5 |
|---|---|---|
| SWE-bench Verified (bash only) | 66.6% | 70.8% |
| LMArena WebDev | 1330 | 1437 |
| SWE-bench Multilingual | 64.7% | 67.3% |
| SciCode | 43.3% | 49% |
| WeirdML | 45.4% | 45.6% |
| LMArena Coding | 1453 | 1474 |
| ALE-Bench | 653.48 | 821.65 |
| SWE-bench Verified | — | 73.8% |

## Agentic & Tool Use

- Claude Haiku 4.5: 33.6 (#52)
- Kimi K2.5: 34.2 (#48)

| Benchmark | Claude Haiku 4.5 | Kimi K2.5 |
|---|---|---|
| Terminal-Bench | 35.5% | 43.2% |
| Vending-Bench 2 | 458.89 | 1,198 |
| Berkeley Function Calling Leaderboard | 68.7% | — |
| DeepResearch Bench | 45.5% | — |
| OSWorld | — | 63.3% |
| BALROG | 31.2% | — |
| ExploitBench | 13.7% | — |

## Reasoning

- Claude Haiku 4.5: 15.1 (#320)
- Kimi K2.5: 31.2 (#80)

| Benchmark | Claude Haiku 4.5 | Kimi K2.5 |
|---|---|---|
| ARC-AGI-2 | 4% | 11.8% |
| NYT Connections (extended) | 14.3% | 69.9% |
| ARC-AGI-1 | 47.7% | 65.3% |
| CritPt | 0% | 3.1% |
| Chess Puzzles | 8% | 12% |
| LMArena Hard Prompts | 1420 | 1453 |
| Epoch Capabilities Index | 142.41 | 148.03 |
| SimpleBench | — | 46.8% |
| Kagi LLM Benchmark | — | 78.5% |
| EnigmaEval | — | 3.4% |
| Thematic Generalization | — | 69.4% |
| DTBench | 73.6% | — |
| LMCA | 30.9% | — |
| ForecastBench | 61.4 | — |

## Math

- Claude Haiku 4.5: 44.9 (#78)
- Kimi K2.5: 51.8 (#53)

| Benchmark | Claude Haiku 4.5 | Kimi K2.5 |
|---|---|---|
| OTIS Mock AIME 2024-2025 | 66.7% | 92.2% |
| LMArena Math | 1396 | 1470 |
| FrontierMath (Feb 2025 set) | 5.9% | 27.9% |
| FrontierMath Tier 4 (v1) | 2.1% | 4.2% |
| MathArena Final-Answer Competitions | — | 62.3% |
| Omni-MATH | 56.1% | — |
| MATH Level 5 | 96.4% | — |

## Knowledge

- Claude Haiku 4.5: 37.7 (#153)
- Kimi K2.5: 53.6 (#56)

| Benchmark | Claude Haiku 4.5 | Kimi K2.5 |
|---|---|---|
| GPQA Diamond | 71.2% | 87.6% |
| SimpleQA Verified | 13.2% | 34.3% |
| Vectara Hallucination Rate | 9.8% | 14.2% |
| LMArena Expert | 1442 | 1466 |
| Humanity's Last Exam | — | 24.4% |
| MMLU-Pro | 77.7% | — |
| GPQA (HELM) | 60.5% | — |

## Multimodal

- Claude Haiku 4.5: 26.8 (#118)
- Kimi K2.5: 41.1 (#39)

| Benchmark | Claude Haiku 4.5 | Kimi K2.5 |
|---|---|---|
| LMArena Document | 1420 | 1430 |
| LMArena Vision | — | 1269 |
| Blueprint-Bench 2 | 0% | — |

## Multilingual

- Claude Haiku 4.5: 49.9 (#129)
- Kimi K2.5: 53.9 (#53)

| Benchmark | Claude Haiku 4.5 | Kimi K2.5 |
|---|---|---|
| LMArena Non-English | 1377 | 1433 |
| LMArena Chinese | 1417 | 1495 |
| LMArena French | 1408 | 1454 |
| LMArena German | 1375 | 1441 |
| LMArena Japanese | 1339 | 1421 |
| LMArena Korean | 1347 | 1410 |
| LMArena Russian | 1381 | 1435 |
| LMArena Spanish | 1420 | 1450 |

## Instruction Following

- Claude Haiku 4.5: 71.4 (#149)
- Kimi K2.5: 75.3 (#64)

| Benchmark | Claude Haiku 4.5 | Kimi K2.5 |
|---|---|---|
| LMArena Instruction Following | 1414 | 1431 |
| IFEval | 80.1% | — |

## Long Context

- Claude Haiku 4.5: 43.6 (#92)
- Kimi K2.5: 52.1 (#7)

| Benchmark | Claude Haiku 4.5 | Kimi K2.5 |
|---|---|---|
| LMArena Longer Query | 1427 | 1445 |
| Fiction.LiveBench | — | 86.1% |
| CL-bench | — | 19.3% |
| CL-bench Life | — | 13.2% |

## Writing & Preference

- Claude Haiku 4.5: 57.9 (#123)
- Kimi K2.5: 65.1 (#53)

| Benchmark | Claude Haiku 4.5 | Kimi K2.5 |
|---|---|---|
| LMArena Text | 1396 | 1445 |
| LMArena Creative Writing | 1372 | 1423 |
| LMArena Multi-Turn | 1409 | 1444 |
| EQ-Bench Creative Writing | — | 1579 |
| WildBench | 83.9% | — |
| EQ-Bench 4 | 1064 | — |

## FAQ

### Is Claude Haiku 4.5 better than Kimi K2.5?

Kimi K2.5 is the stronger model overall, scoring 48.1 to 39.5 on the Noometry Index.

### Which is cheaper, Claude Haiku 4.5 or Kimi K2.5?

Kimi K2.5 is cheaper. It lists at $0.45 per million input tokens and $2.25 per million output tokens; Claude Haiku 4.5 lists at $1 and $5.

### Is Claude Haiku 4.5 or Kimi K2.5 better for coding?

Kimi K2.5 scores higher on coding benchmarks: 48.8 versus 44.0 in the Noometry coding category.

### Which has the bigger context window?

Kimi K2.5 does, with 262K tokens against 200K.

### How many benchmarks do Claude Haiku 4.5 and Kimi K2.5 share?

38 benchmarks have published results for both models. Claude Haiku 4.5 has 53 scored results on Noometry and Kimi K2.5 has 51.