Side-by-side benchmark comparison across agentic, coding, multimodal, knowledge, reasoning, and math workflows.
Kimi K2.5 (Reasoning)
76
Winner · 5/8 categoriesQwen3.5-122B-A10B
71
2/8 categoriesKimi K2.5 (Reasoning)· Qwen3.5-122B-A10B
Pick Kimi K2.5 (Reasoning) if you want the stronger benchmark profile. Qwen3.5-122B-A10B only becomes the better choice if knowledge is the priority or you need the larger 262K context window.
Kimi K2.5 (Reasoning) is clearly ahead on the aggregate, 76 to 71. The gap is large enough that you do not need to squint at the spreadsheet to see the difference.
Kimi K2.5 (Reasoning)'s sharpest advantage is in reasoning, where it averages 74.3 against 60.2. The single biggest benchmark swing on the page is SuperGPQA, 88% to 67.1%. Qwen3.5-122B-A10B does hit back in knowledge, so the answer changes if that is the part of the workload you care about most.
Qwen3.5-122B-A10B gives you the larger context window at 262K, compared with 128K for Kimi K2.5 (Reasoning).
BenchLM keeps the benchmark table and the operator tradeoffs on the same page so a better score does not hide a materially slower, pricier, or smaller-context model.
Runtime metrics show N/A when BenchLM does not have a sourced snapshot for that exact model. The scoring rules and freshness policy are documented on the methodology page.
| Benchmark | Kimi K2.5 (Reasoning) | Qwen3.5-122B-A10B |
|---|---|---|
| AgenticKimi K2.5 (Reasoning) wins | ||
| Terminal-Bench 2.0 | 50.8% | 49.4% |
| BrowseComp | 60.6% | 63.8% |
| OSWorld-Verified | 63.3% | 58% |
| tau2-bench | — | 79.5% |
| CodingQwen3.5-122B-A10B wins | ||
| HumanEval | 99% | — |
| SWE-bench Verified | 76.8% | 72% |
| LiveCodeBench | 85% | 78.9% |
| SWE-bench Pro | 70% | — |
| SWE-Rebench | 57.4% | — |
| Multimodal & GroundedKimi K2.5 (Reasoning) wins | ||
| MMMU-Pro | 78.5% | 76.9% |
| OfficeQA Pro | 77% | — |
| ReasoningKimi K2.5 (Reasoning) wins | ||
| MuSR | 86% | — |
| BBH | 91% | — |
| LongBench v2 | 61% | 60.2% |
| MRCRv2 | 81% | — |
| KnowledgeQwen3.5-122B-A10B wins | ||
| MMLU | 92% | — |
| GPQA | 87.6% | 86.6% |
| SuperGPQA | 88% | 67.1% |
| MMLU-Pro | 87.1% | 86.7% |
| HLE | 27% | — |
| FrontierScience | 80% | — |
| SimpleQA | 54% | — |
| Instruction FollowingKimi K2.5 (Reasoning) wins | ||
| IFEval | 94% | 93.4% |
| MultilingualKimi K2.5 (Reasoning) wins | ||
| MGSM | 96% | — |
| MMLU-ProX | 86% | 82.2% |
| Mathematics | ||
| AIME 2023 | 94% | — |
| AIME 2024 | 96% | — |
| AIME 2025 | 96.1% | — |
| HMMT Feb 2023 | 90% | — |
| HMMT Feb 2024 | 92% | — |
| HMMT Feb 2025 | 95.4% | — |
| BRUMO 2025 | 93% | — |
| MATH-500 | 92% | — |
Kimi K2.5 (Reasoning) is ahead overall, 76 to 71. The biggest single separator in this matchup is SuperGPQA, where the scores are 88% and 67.1%.
Qwen3.5-122B-A10B has the edge for knowledge tasks in this comparison, averaging 81.6 versus 67.9. Inside this category, SuperGPQA is the benchmark that creates the most daylight between them.
Qwen3.5-122B-A10B has the edge for coding in this comparison, averaging 76.3 versus 70.4. Inside this category, LiveCodeBench is the benchmark that creates the most daylight between them.
Kimi K2.5 (Reasoning) has the edge for reasoning in this comparison, averaging 74.3 versus 60.2. Inside this category, LongBench v2 is the benchmark that creates the most daylight between them.
Kimi K2.5 (Reasoning) has the edge for agentic tasks in this comparison, averaging 57.6 versus 56. Inside this category, OSWorld-Verified is the benchmark that creates the most daylight between them.
Kimi K2.5 (Reasoning) has the edge for multimodal and grounded tasks in this comparison, averaging 77.8 versus 76.9. Inside this category, MMMU-Pro is the benchmark that creates the most daylight between them.
Kimi K2.5 (Reasoning) has the edge for instruction following in this comparison, averaging 94 versus 93.4. Inside this category, IFEval is the benchmark that creates the most daylight between them.
Kimi K2.5 (Reasoning) has the edge for multilingual tasks in this comparison, averaging 89.5 versus 82.2. Inside this category, MMLU-ProX is the benchmark that creates the most daylight between them.
Get notified when new models drop, benchmark scores change, or the leaderboard shifts. One email per week.
Free. No spam. Unsubscribe anytime. We only store derived location metadata for consent routing.