Qwen 3.6
#3alibaba · v3.6 · seit April 2026 · 160× · zuletzt 30. Juni 2026
72
Momentum
Qwen 3.6 is a language model by Alibaba available in multiple sizes (27B, 35B). It is a Mixture-of-Experts agentic model equipped with multi-token prediction and achieves GPT-4-class performance. Parts of the model have been released as open-source variants.
Historique du momentum
04.04.03.07.
Fonctionnalités
| Key Benchmark (%) | Qwen3.6-35B-A3B: SWE-Bench Verified 73.4%; Terminal-Bench 2.0: 51.5%; AIME 2026: 92.6%. Qwen3.6-Plus: SWE-Bench Verified 78.8% |
| Context Window (Tokens) | Qwen3.6-35B-A3B: 262,144 native, extendable to 1,010,000 (via YaRN); Qwen3.6-Plus: 1,000,000 native; Qwen3.6-Max-Preview: 256,000 |
| License | Qwen3.6-35B-A3B: Apache 2.0 (open weights, commercially usable); Qwen3.6-Plus / Max-Preview: proprietary, API-only |
| Multimodality | Yes – text, image, video as input (natively multimodal); Qwen3.6-Max-Preview is text-only (no image input) |
| Platform | Alibaba Cloud Model Studio (API); Qwen Studio (chat UI); Hugging Face & ModelScope (open weights); OpenRouter, Together AI, DeepInfra (third-party) |
| Price | Qwen3.6-Plus: $0.325/1M input, $1.95/1M output (OpenRouter); Qwen3.6-35B-A3B: $0.14/1M input, $1.00/1M output (OpenRouter) |
| Price per 1M Tokens | Qwen3.6-Plus: $0.325 input / $1.95 output; Qwen3.6-35B-A3B: $0.14 input / $1.00 output (via OpenRouter/Alibaba Cloud) |
| Release Date | Qwen3.6-Plus: April 2, 2026 (API); Qwen3.6-35B-A3B open weights: April 16, 2026; Qwen3.6-Max-Preview: April 20, 2026 |
