Model comparison
Kimi K2.5 vs Qwen3 235B-A22B
Kimi K2.5 is the stronger model overall, scoring 47.8 to 43.2 on the Noometry Index.
Last verified . 35 shared benchmarks.
Alibaba (Qwen)Summary
- They share 35 benchmarks with published results for both. Kimi K2.5 scores higher in 9 categories and Qwen3 235B-A22B in 0 categories; 8 gaps are clear of the uncertainty.
- The widest gap is in reasoning, where Kimi K2.5 leads 31.2 to 15.8.
- The biggest single-benchmark swing is ARC-AGI-1: 65.3% for Kimi K2.5 and 11% for Qwen3 235B-A22B.
- Kimi K2.5 is cheaper at $0.45 / $2.25 per million input/output tokens, against $0.70 / $2.80 for Qwen3 235B-A22B.
- Kimi K2.5 accepts more context: 262K tokens versus 131K.
Side by side
| Kimi K2.5 | Qwen3 235B-A22B | |
|---|---|---|
| Provider | Moonshot AI | Alibaba (Qwen) |
| Noometry Index | 47.8 | 43.2 |
| Released | 2026-01-27 | 2025-04 |
| Weights | Open | Open |
| Context window | 262K | 131K |
| Max output | 262K | 16K |
| Input $ / M tokens | $0.45 | $0.70 |
| Output $ / M tokens | $2.25 | $2.80 |
| Results tracked | 51 | 49 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding Kimi K2.5 leads
Kimi K2.5: 48.7 (#48), Qwen3 235B-A22B: 43.8 (#72)
| Benchmark | Kimi K2.5 | Qwen3 235B-A22B |
|---|---|---|
| SciCode | 49% | 42.4% |
| WeirdML | 45.6% | 41% |
| LMArena Coding | 1474 | 1445 |
| SWE-bench Verified | 73.8% | — |
| SWE-bench Verified (bash only) | 70.8% | — |
| Aider Polyglot | — | 59.6% |
| LMArena WebDev | 1436 | — |
| SWE-bench Multilingual | 67.3% | — |
| ALE-Bench | 821.65 | — |
Agentic & Tool Use Too close to call
Kimi K2.5: 34.0 (#44), Qwen3 235B-A22B: 33.7 (#48)
| Benchmark | Kimi K2.5 | Qwen3 235B-A22B |
|---|---|---|
| Vending-Bench 2 | 1,198 | -11.34 |
| Terminal-Bench | 43.2% | — |
| Berkeley Function Calling Leaderboard | — | 52.1% |
| OSWorld | 63.3% | — |
Reasoning Kimi K2.5 leads
Kimi K2.5: 31.2 (#74), Qwen3 235B-A22B: 15.8 (#302)
| Benchmark | Kimi K2.5 | Qwen3 235B-A22B |
|---|---|---|
| ARC-AGI-2 | 11.8% | 1.3% |
| SimpleBench | 46.8% | 31% |
| Kagi LLM Benchmark | 78.5% | 69.4% |
| ARC-AGI-1 | 65.3% | 11% |
| CritPt | 3.1% | 0% |
| Chess Puzzles | 12% | 12% |
| LMArena Hard Prompts | 1454 | 1433 |
| Epoch Capabilities Index | 148.05 | 143.87 |
| NYT Connections (extended) | 69.9% | — |
| EnigmaEval | 3.4% | — |
| Thematic Generalization | 69.4% | — |
| Mystery Game Puzzles | — | 9% |
| DTBench | — | 80.3% |
| LMCA | — | 29.3% |
| ForecastBench | — | 59.7 |
Math Kimi K2.5 leads
Kimi K2.5: 50.3 (#46), Qwen3 235B-A22B: 48.5 (#51)
| Benchmark | Kimi K2.5 | Qwen3 235B-A22B |
|---|---|---|
| OTIS Mock AIME 2024-2025 | 92.2% | 86.7% |
| LMArena Math | 1470 | 1431 |
| FrontierMath (Feb 2025 set) | 27.9% | 8.5% |
| FrontierMath Tier 4 (v1) | 4.2% | 0% |
| MathArena Final-Answer Competitions | 62.3% | — |
| Omni-MATH | — | 71.8% |
| MATH Level 5 | — | 68.9% |
Knowledge Kimi K2.5 leads
Kimi K2.5: 53.6 (#52), Qwen3 235B-A22B: 49.5 (#67)
| Benchmark | Kimi K2.5 | Qwen3 235B-A22B |
|---|---|---|
| GPQA Diamond | 87.6% | 80.1% |
| SimpleQA Verified | 34.3% | 40.4% |
| Vectara Hallucination Rate | 14.2% | 9.3% |
| LMArena Expert | 1466 | 1460 |
| Humanity's Last Exam | 24.4% | — |
| MMLU-Pro | — | 84.4% |
| Confabulations | — | 15.6% |
| GPQA (HELM) | — | 72.7% |
Multimodal Not comparable
Kimi K2.5: 40.4 (#37), Qwen3 235B-A22B: —
| Benchmark | Kimi K2.5 | Qwen3 235B-A22B |
|---|---|---|
| LMArena Vision | 1269 | — |
| LMArena Document | 1430 | — |
Multilingual Kimi K2.5 leads
Kimi K2.5: 54.1 (#50), Qwen3 235B-A22B: 52.5 (#84)
| Benchmark | Kimi K2.5 | Qwen3 235B-A22B |
|---|---|---|
| LMArena Non-English | 1433 | 1410 |
| LMArena Chinese | 1495 | 1481 |
| LMArena French | 1454 | 1445 |
| LMArena German | 1441 | 1432 |
| LMArena Japanese | 1422 | 1400 |
| LMArena Korean | 1412 | 1391 |
| LMArena Russian | 1436 | 1411 |
| LMArena Spanish | 1450 | 1429 |
Instruction Following Kimi K2.5 leads
Kimi K2.5: 75.5 (#60), Qwen3 235B-A22B: 72.8 (#132)
| Benchmark | Kimi K2.5 | Qwen3 235B-A22B |
|---|---|---|
| LMArena Instruction Following | 1431 | 1408 |
| IFEval | — | 83.5% |
Long Context Kimi K2.5 leads
Kimi K2.5: 52.1 (#7), Qwen3 235B-A22B: 46.1 (#25)
| Benchmark | Kimi K2.5 | Qwen3 235B-A22B |
|---|---|---|
| Fiction.LiveBench | 86.1% | 75% |
| LMArena Longer Query | 1445 | 1426 |
| CL-bench | 19.3% | — |
| CL-bench Life | 13.2% | — |
Writing & Preference Kimi K2.5 leads
Kimi K2.5: 65.3 (#51), Qwen3 235B-A22B: 59.8 (#103)
| Benchmark | Kimi K2.5 | Qwen3 235B-A22B |
|---|---|---|
| LMArena Text | 1445 | 1419 |
| LMArena Creative Writing | 1423 | 1384 |
| EQ-Bench Creative Writing | 1579 | 1366 |
| LMArena Multi-Turn | 1444 | 1432 |
| Short-Story Creative Writing | — | 83% |
| WildBench | — | 86.6% |
Frequently asked questions
Is Kimi K2.5 better than Qwen3 235B-A22B?
Kimi K2.5 is the stronger model overall, scoring 47.8 to 43.2 on the Noometry Index.
Which is cheaper, Kimi K2.5 or Qwen3 235B-A22B?
Kimi K2.5 is cheaper. It lists at $0.45 per million input tokens and $2.25 per million output tokens; Qwen3 235B-A22B lists at $0.70 and $2.80.
Is Kimi K2.5 or Qwen3 235B-A22B better for coding?
Kimi K2.5 scores higher on coding benchmarks: 48.7 versus 43.8 in the Noometry coding category.
Which has the bigger context window?
Kimi K2.5 does, with 262K tokens against 131K.
How many benchmarks do Kimi K2.5 and Qwen3 235B-A22B share?
35 benchmarks have published results for both models. Kimi K2.5 has 51 scored results on Noometry and Qwen3 235B-A22B has 49.