Up to four models
Kimi K2.6 vs Qwen3.6-Plus
On the index (GPT-4 = 100), Kimi K2.6 scores 152 and Qwen3.6-Plus 145. Side by side on the same third-party tests, with price and context. The link in your address bar shares this exact comparison.
Kimi K2.6
152136–169
WeightsOpen
ReleasedApr 2026
Price in / out$0.65 / $3.41
Context262K
Core tests taken4 of 9
145128–162
WeightsClosed
ReleasedApr 2026
Price in / out$0.33 / $1.95
Context1M
Core tests taken4 of 9
Kimi K2.6 and Qwen3.6-Plus are too close to call (7 points apart, within the uncertainty), ahead on 4 of 4 shared core tests, at 1.8× the price.
Test by testthird-party results
Kimi K2.6
Qwen3.6-Plus
Every result as a table
| Test | Kimi K2.6 | Qwen3.6-Plus |
|---|---|---|
| Reasoning & knowledge | ||
| GPQA Diamond (Epoch AI) | 90.8% | 88.4% |
| LiveBench Data Analysis | 65.1% | 69.9% |
| LiveBench IF | 64.4% | 58.3% |
| LiveBench Language | 75.1% | 75.0% |
| LiveBench Reasoning | 79.4% | 75.8% |
| GPQA Diamond (Vals AI) | 89.1% | 87.4% |
| MMLU Pro | 87.6% | 87.7% |
| MMMU Pro | 86.3% | 84.2% |
| SimpleQA Verified | 34.9% | 44.1% |
| Maths | ||
| FrontierMath (tiers 1-3) | 57.2% | 38.3% |
| AIME-style maths (OTIS mock) | 96.1% | 93.3% |
| LiveBench Mathematics | 84.3% | 83.7% |
| Chess puzzles | 20.0% | 16.0% |
| FrontierMath (2025 set) | 39.0% | 26.2% |
| FrontierMath tier 4 (2025 set) | 14.6% | 8.3% |
| Coding | ||
| SWE-bench Verified | 76.6% | 57.9% |
| LiveBench Agentic Coding | 46.9% | 41.4% |
| LiveBench Coding | 78.6% | 78.2% |
| Code Migration | 27.8% | 11.1% |
| LiveCodeBench | 86.8% | 86.0% |
| ProgramBench | 0.0% | 0.0% |
| SWE-bench | 76.2% | 73.4% |
| Vibe Code Bench v1.1 | 37.9% | 25.6% |
| Agents | ||
| DTBench | 90.9% | 81.9% |
| LMCA | 37.3% | 33.1% |
| Mystery game puzzles | 18.0% | 12.0% |
| Professional work | ||
| Finance Agent (v2) | 44.9% | 40.8% |
| CorpFin v2 | 66.7% | 61.9% |
| EMB | 57.9% | 32.9% |
| LegalBench | 84.7% | 84.2% |
| Legal Research Bench | 15.9% | 14.9% |
| MedCode | 40.1% | 36.9% |
| MedScribe | 78.1% | 77.0% |
| MortgageTax | 65.8% | 68.0% |
| Tax Agent Bench | 12.4% | 7.7% |
| TaxEval v2 | 74.7% | 74.7% |
Tests at least two of these models have taken. Bold rows are the core tests.