Qwen3.5-27B

Alibaba / Qwen🇨🇳 China
active
Context window262K tokens
Input / 1M tokens$0.26
Output / 1M tokens$2.6

Version History

3.5minor

Qwen3.5-27B released as a 27-billion parameter multimodal model supporting image-text-to-text tasks. Available under Apache 2.0 license with transformer endpoint compatibility.

Benchmark Scores

Full leaderboard →
1408.0 elo
Arena Elo
85.5%
GPQA
12.1%
Hallucination (Summarization)
74.1%
Hallucination Rate
86.1%
MMLU-Pro
82.3%
MMMU
82.0 tokens_per_sec
Speed (tok/s)
72.4%
SWE-bench Verified

Coverage