Qwen3.5 397B A17B
Alibaba / Qwen🇨🇳 China
Qwen3.5 series 397B-A17B native vision-language model built on a hybrid architecture integrating linear attention with a sparse mixture-of-experts model. State-of-the-art performance with higher inference efficiency.
Context window262K tokens
Input / 1M tokens$0.39
Output / 1M tokens$2.34
Version History
qwen3.5-397b-a17b-2026-02-16major
Qwen3.5 397B A17B launches as a hybrid linear-attention + sparse MoE vision-language model. 262K context at $0.39/$2.34 per 1M tokens with state-of-the-art performance.
Benchmark Scores
Full leaderboard →91.3%
AIME 2026
1442.0 elo
Arena Elo
89.3%
GPQA
82.7%
Hallucination Rate
87.8%
MMLU-Pro
78.0 tokens_per_sec
Speed (tok/s)
76.4%
SWE-bench Verified