Qwen3.5-9B

Qwen · released 2026-03-01 · Apache-2.0 license

Qwen3.5-9B is Alibaba's small, natively multimodal reasoning model (March 2026, Apache-2.0) with 262K context, scoring 32 on the Artificial Analysis Intelligence Index, well above the median for its size.

Key specs

TypeLocal open-weight
Parameters
Architecture
Context window
Knowledge cutoff
Modalitiestext, vision
Recommended backends
Minimum viable rig

Benchmark scores

GPQA Diamond81.7%
SWE-bench Verified
AIME92.5%
MMLU-Pro82.5%
BFCL v3 (tool use)
Composite score8.56
Community ratingNo reviews yet

Strengths & weaknesses

Strengths: Very high intelligence-per-size (AA Index 32); Natively multimodal at 9B with 262K context; Strong instruction following (IFEval 91.5) and tool use (BFCL-V4 66.1); Apache-2.0

Weaknesses: Slow, verbose output for its size; Coding (LiveCodeBench 65.6) trails larger MoE peers