Model comparison
Kimi K2 vs Qwen3.6-27B
Head-to-head evidence from 11 shared benchmark results across 5 categories. Overall scores shown here use the public BenchAlign v5 ranking lane.
Public leaderboard positions: Kimi K2 #189 (Supported); Qwen3.6-27B #93 (Estimated). Intervals and evidence labels describe ranking uncertainty, not a guarantee for a specific workload.
Evidence parity. Kimi K2 and Qwen3.6-27B share 11 comparable benchmark results. 1 of 8 categories are comparable. 3 results are unique to Kimi K2; 43 to Qwen3.6-27B.
Updated July 21, 2026- Shared results
- 11
- Kimi K2 only
- 3
- Qwen3.6-27B only
- 43
- Comparable categories
- 1 / 8
Pick Qwen3.6-27B if you want the stronger benchmark profile. Kimi K2 only becomes the better choice if you would rather avoid the extra latency and token burn of a reasoning model.
Confidence note. This is a partial-evidence comparison with 11 shared benchmark results across 5 evidence categories; 1 of 8 categories currently have scoreable aggregates for both models. Treat the verdict as directional until coverage is more balanced.
Why this result
Qwen3.6-27B is clearly ahead on the BenchAlign aggregate, 53.82 to 27.19. The gap is large enough that you do not need to squint at the spreadsheet to see the difference.
Qwen3.6-27B's sharpest advantage is in mathematics, where it averages 89.2 against 16.1.
Kimi K2 is also the more expensive model on tokens at $0.60 input / $2.50 output per 1M tokens, versus $0.00 input / $0.00 output per 1M tokens for Qwen3.6-27B. That is roughly Infinityx on output cost alone. Qwen3.6-27B is the reasoning model in the pair, while Kimi K2 is not. That usually helps on harder chain-of-thought-heavy tests, but it can also mean more latency and more token spend in real use. Qwen3.6-27B gives you the larger context window at 262K, compared with 128K for Kimi K2.
Category breakdown
Exact category averages are shown below. Not measured means BenchLM does not have enough sourced public coverage for that model and category.
| Category | Kimi K2 | Δ | Qwen3.6-27B |
|---|---|---|---|
| Math | Kimi K216.1 | Margin→ 73.1 | Qwen3.6-27B89.2 |
| Agentic | Kimi K2Not measured | MarginNo overlap | Qwen3.6-27B59.3 |
| Coding | Kimi K2Not measured | MarginNo overlap | Qwen3.6-27B77.5 |
| Knowledge | Kimi K2Not measured | MarginNo overlap | Qwen3.6-27B53.3 |
| Multimodal | Kimi K2Not measured | MarginNo overlap | Qwen3.6-27B76.7 |
Operational comparison
Runtime and commercial metrics are compared only when both models have a complete sourced value.
| Metric | Kimi K2 | Qwen3.6-27B | Comparison |
|---|---|---|---|
| Input / output priceUSD per 1M tokens | Kimi K2$0.6 input / $2.5 output | Qwen3.6-27B$0 input / $0 output | Qwen3.6-27B has the lower combined listed price. |
| Generation speedtokens per second | Kimi K243 tok/s | Qwen3.6-27BNot available | A complete speed comparison is not available. |
| First-answer latencyseconds to first token | Kimi K21.51 s | Qwen3.6-27BNot available | A complete latency comparison is not available. |
| Context windowmaximum listed tokens | Kimi K2128K | Qwen3.6-27B262K | Qwen3.6-27B lists the larger context window. |
Benchmark Deep Dive
Agentic10 benchmarks
| Benchmark | Kimi K2 | Qwen3.6-27B | Result |
|---|---|---|---|
| τ²-bench resultsSource | 61.1% | 94.2% | Qwen3.6-27B leads |
| Terminal-Bench 2.0Source | — | 59.3% | Not comparable |
| Claw-EvalSource | — | 72.4% | Not comparable |
| QwenClawBenchSource | — | 53.4% | Not comparable |
| QwenWebBenchSource | — | 1487 | Not comparable |
| AndroidWorldSource | — | 70.3% | Not comparable |
| AA Agentic IndexSource | — | 27.0% | Not comparable |
| GDPval-AASource | — | 32.0% | Not comparable |
| GDPval-AASource | — | 1140 | Not comparable |
| Gert LabsSource | — | 54.84% | Not comparable |
Coding8 benchmarks
| Benchmark | Kimi K2 | Qwen3.6-27B | Result |
|---|---|---|---|
| AA-SciCodeSource | 34.5% | 39.8% | Qwen3.6-27B leads |
| SWE-bench VerifiedSource | — | 77.2% | Not comparable |
| SWE MultilingualSource | — | 71.3% | Not comparable |
| SWE-bench ProSource | — | 53.5% | Not comparable |
| Terminal-Bench 2.0Source | — | 59.3% | Not comparable |
| LiveCodeBenchSource | — | 83.9% | Not comparable |
| NL2RepoSource | — | 36.2% | Not comparable |
| AA Coding IndexSource | — | 53.7% | Not comparable |
Reasoning2 benchmarks
Knowledge12 benchmarks
| Benchmark | Kimi K2 | Qwen3.6-27B | Result |
|---|---|---|---|
| Artificial Analysis Intelligence IndexSource | 19.4% | 37.0% | Qwen3.6-27B leads |
| AA-GPQA DiamondSource | 76.6% | 84.2% | Qwen3.6-27B leads |
| AA-HLESource | 7.0% | 21.6% | Qwen3.6-27B leads |
| AA-Omniscience IndexSource | -27.5% | -19.8% | Qwen3.6-27B leads |
| AA-Omniscience AccuracySource | 26.8% | 19.2% | Kimi K2 leads |
| AA-Omniscience Hallucination RateSource | 74.2% | 48.3% | Qwen3.6-27B leads |
| MMLU-ProSource | — | 86.2% | Not comparable |
| MMLU-ReduxSource | — | 93.5% | Not comparable |
| SuperGPQASource | — | 66% | Not comparable |
| C-EvalSource | — | 91.4% | Not comparable |
| GPQASource | — | 87.8% | Not comparable |
| HLESource | — | 24% | Not comparable |
MathQwen3.6-27B wins7 benchmarks
| Benchmark | Kimi K2 | Qwen3.6-27B | Result |
|---|---|---|---|
| FrontierMath v2 (Tiers 1-3)Source | 21.404% | — | Not comparable |
| FrontierMath v2 (Tier 4)Source | 0.000% | — | Not comparable |
| HMMT Feb 2025Source | — | 93.8% | Not comparable |
| HMMT Nov 2025Source | — | 90.7% | Not comparable |
| HMMT Feb 2026Source | — | 84.3% | Not comparable |
| MMAnswerBenchSource | — | 80.8% | Not comparable |
| AIME26Source | — | 94.1% | Not comparable |
Multimodal17 benchmarks
| Benchmark | Kimi K2 | Qwen3.6-27B | Result |
|---|---|---|---|
| Design Arena WebsiteSource | 1080 | — | Not comparable |
| MMMUSource | — | 82.9% | Not comparable |
| MMMU-ProSource | — | 75.8% | Not comparable |
| RealWorldQASource | — | 84.1% | Not comparable |
| DynaMathSource | — | 85.6% | Not comparable |
| MStarSource | — | 81.4% | Not comparable |
| SimpleVQASource | — | 56.1% | Not comparable |
| CharXivSource | — | 78.4% | Not comparable |
| CC-OCRSource | — | 81.2% | Not comparable |
| CountBenchSource | — | 97.8% | Not comparable |
| RefCOCO (avg)Source | — | 92.5% | Not comparable |
| ERQASource | — | 62.5% | Not comparable |
| Video-MME (with subtitle)Source | — | 87.7% | Not comparable |
| VideoMMMUSource | — | 84.4% | Not comparable |
| MLVU (M-Avg)Source | — | 86.6% | Not comparable |
| V*Source | — | 94.7% | Not comparable |
| AA-MMMU-ProSource | — | 74.6% | Not comparable |
Inst. Following1 benchmarks
| Benchmark | Kimi K2 | Qwen3.6-27B | Result |
|---|---|---|---|
| AA-IFBenchSource | 41.5% | 67.6% | Qwen3.6-27B leads |
Frequently Asked Questions (2)
Which is better, Kimi K2 or Qwen3.6-27B?
Qwen3.6-27B is ahead on BenchLM's BenchAlign leaderboard, 53.82 to 27.19.
Which is better for math, Kimi K2 or Qwen3.6-27B?
Qwen3.6-27B has the edge for math in this comparison, averaging 89.2 versus 16.1. Kimi K2 stays close enough that the answer can still flip depending on your workload.
Self-host vs API cost
Estimates at 50,000 req/day · 1000 tokens/req average.
Related Comparisons
Choose a model with this week’s evidence
Join 2,000+ readers for ranking moves, pricing changes, and the claims that still need proof.
One email each week. Unsubscribe anytime.