Model comparison
Kimi K2.5 vs Qwen3.8 27B
Kimi K2.5 is the stronger model overall, scoring 48.1 to 46.0 on the Noometry Index.
Last verified . 26 shared benchmarks.
Alibaba (Qwen)Summary
- They share 26 benchmarks with published results for both. Kimi K2.5 scores higher in 5 categories and Qwen3.8 27B in 5 categories; 6 gaps are clear of the uncertainty.
- The widest gap is in math, where Kimi K2.5 leads 51.8 to 37.1.
- The biggest single-benchmark swing is ARC-AGI-2: 11.8% for Kimi K2.5 and 42.4% for Qwen3.8 27B.
- Kimi K2.5 is cheaper at $0.45 / $2.25 per million input/output tokens, against $0.99 / $1.49 for Qwen3.8 27B.
Side by side
| Kimi K2.5 | Qwen3.8 27B | |
|---|---|---|
| Provider | Moonshot AI | Alibaba (Qwen) |
| Noometry Index | 48.1 | 46.0 |
| Released | 2026-01-27 | 2026-08-14 |
| Weights | Open | Open |
| Context window | 262K | 262K |
| Max output | 262K | 33K |
| Input $ / M tokens | $0.45 | $0.99 |
| Output $ / M tokens | $2.25 | $1.49 |
| Results tracked | 51 | 31 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding Qwen3.8 27B leads
Kimi K2.5: 48.8 (#53), Qwen3.8 27B: 50.5 (#44)
| Benchmark | Kimi K2.5 | Qwen3.8 27B |
|---|---|---|
| LMArena WebDev | 1437 | 1593 |
| SciCode | 49% | 46.6% |
| LMArena Coding | 1474 | 1482 |
| SWE-bench Verified | 73.8% | — |
| SWE-bench Verified (bash only) | 70.8% | — |
| SWE-bench Multilingual | 67.3% | — |
| WeirdML | 45.6% | — |
| ALE-Bench | 821.65 | — |
Agentic & Tool Use Kimi K2.5 leads
Kimi K2.5: 34.2 (#48), Qwen3.8 27B: 32.9 (#57)
| Benchmark | Kimi K2.5 | Qwen3.8 27B |
|---|---|---|
| Terminal-Bench | 43.2% | — |
| APEX-Agents | — | 47.5% |
| OSWorld | 63.3% | — |
| Vending-Bench 2 | 1,198 | — |
Reasoning Qwen3.8 27B leads
Kimi K2.5: 31.2 (#80), Qwen3.8 27B: 41.0 (#54)
| Benchmark | Kimi K2.5 | Qwen3.8 27B |
|---|---|---|
| ARC-AGI-2 | 11.8% | 42.4% |
| NYT Connections (extended) | 69.9% | 54.5% |
| ARC-AGI-1 | 65.3% | 87.5% |
| CritPt | 3.1% | 5.4% |
| LMArena Hard Prompts | 1453 | 1460 |
| Epoch Capabilities Index | 148.03 | 149.38 |
| SimpleBench | 46.8% | — |
| Kagi LLM Benchmark | 78.5% | — |
| Chess Puzzles | 12% | — |
| EnigmaEval | 3.4% | — |
| Thematic Generalization | 69.4% | — |
| DTBench | — | 88% |
| LMCA | — | 41.4% |
| Surface Evolver Bench | — | 45% |
Math Kimi K2.5 leads
Kimi K2.5: 51.8 (#53), Qwen3.8 27B: 37.1 (#161)
| Benchmark | Kimi K2.5 | Qwen3.8 27B |
|---|---|---|
| LMArena Math | 1470 | 1456 |
| MathArena Final-Answer Competitions | 62.3% | — |
| OTIS Mock AIME 2024-2025 | 92.2% | — |
| ProofBench | — | 16% |
| FrontierMath (Feb 2025 set) | 27.9% | — |
| FrontierMath Tier 4 (v1) | 4.2% | — |
Knowledge Kimi K2.5 leads
Kimi K2.5: 53.6 (#56), Qwen3.8 27B: 41.6 (#109)
| Benchmark | Kimi K2.5 | Qwen3.8 27B |
|---|---|---|
| LMArena Expert | 1466 | 1482 |
| GPQA Diamond | 87.6% | — |
| Humanity's Last Exam | 24.4% | — |
| SimpleQA Verified | 34.3% | — |
| Vectara Hallucination Rate | 14.2% | — |
Multimodal Too close to call
Kimi K2.5: 41.1 (#39), Qwen3.8 27B: 41.3 (#37)
| Benchmark | Kimi K2.5 | Qwen3.8 27B |
|---|---|---|
| LMArena Vision | 1269 | 1271 |
| LMArena Document | 1430 | — |
Multilingual Too close to call
Kimi K2.5: 53.9 (#53), Qwen3.8 27B: 53.7 (#60)
| Benchmark | Kimi K2.5 | Qwen3.8 27B |
|---|---|---|
| LMArena Non-English | 1433 | 1430 |
| LMArena Chinese | 1495 | 1504 |
| LMArena French | 1454 | 1465 |
| LMArena German | 1441 | 1438 |
| LMArena Japanese | 1421 | 1384 |
| LMArena Korean | 1410 | 1393 |
| LMArena Russian | 1435 | 1415 |
| LMArena Spanish | 1450 | 1448 |
Instruction Following Too close to call
Kimi K2.5: 75.3 (#64), Qwen3.8 27B: 75.8 (#53)
| Benchmark | Kimi K2.5 | Qwen3.8 27B |
|---|---|---|
| LMArena Instruction Following | 1431 | 1439 |
Long Context Kimi K2.5 leads
Kimi K2.5: 52.1 (#7), Qwen3.8 27B: 44.3 (#70)
| Benchmark | Kimi K2.5 | Qwen3.8 27B |
|---|---|---|
| LMArena Longer Query | 1445 | 1450 |
| Fiction.LiveBench | 86.1% | — |
| CL-bench | 19.3% | — |
| CL-bench Life | 13.2% | — |
Writing & Preference Too close to call
Kimi K2.5: 65.1 (#53), Qwen3.8 27B: 65.8 (#43)
| Benchmark | Kimi K2.5 | Qwen3.8 27B |
|---|---|---|
| LMArena Text | 1445 | 1441 |
| LMArena Creative Writing | 1423 | 1384 |
| EQ-Bench Creative Writing | 1579 | 1671 |
| LMArena Multi-Turn | 1444 | 1441 |
Frequently asked questions
Is Kimi K2.5 better than Qwen3.8 27B?
Kimi K2.5 is the stronger model overall, scoring 48.1 to 46.0 on the Noometry Index.
Which is cheaper, Kimi K2.5 or Qwen3.8 27B?
Kimi K2.5 is cheaper. It lists at $0.45 per million input tokens and $2.25 per million output tokens; Qwen3.8 27B lists at $0.99 and $1.49.
Is Kimi K2.5 or Qwen3.8 27B better for coding?
Qwen3.8 27B scores higher on coding benchmarks: 50.5 versus 48.8 in the Noometry coding category.
Which has the bigger context window?
Both accept 262K tokens.
How many benchmarks do Kimi K2.5 and Qwen3.8 27B share?
26 benchmarks have published results for both models. Kimi K2.5 has 51 scored results on Noometry and Qwen3.8 27B has 31.