Qwen3.5 9B

Qwen3.5 9B is a dense, 9-billion-parameter multimodal model built to combine reasoning, coding, and visual understanding in a relatively compact architecture. Its unified vision-language design uses early fusion, so visual tokens are incorporated alongside text rather than handled as a separate pipeline. This makes the model especially relevant for applications that must reason over documents, diagrams, screenshots, or other visual material while retaining language-based problem solving.

The model is designed for tool-oriented and reasoning-enabled workflows, with a native context of 262,144 tokens that supports substantial conversational or project context. Its long-context design and visual-language integration suit coding assistants, document analysis, visual question answering, and agent-like tasks that combine images with instructions. Benchmarks reported for the model show strong graduate-level science reasoning, but more mixed results on long-context and agentic evaluations, so it is best viewed as a balanced compact multimodal option rather than a specialist in every reasoning task.

Mixlayerqwen/qwen3.5-9bqwen

Open provider page Provider docs

Quick Info

Provider: Mixlayer
Model key: qwen/qwen3.5-9b
Release date: Mar 18, 2026
Last updated: Mar 18, 2026
Input modalities
Output modalities
Capabilities

Cost

Input token cost: $0.10
Output token cost: $0.40

Limits

Output tokens: 262,144 tokens
Context window: 262,144 tokens

Transparent token rates

Compare Qwen3.5 9B pricing

Rates are shown per one million tokens. Combined means one million input plus one million output tokens.

Browse this family

Latest news about Qwen3.5 9B

Mixlayer

CoverageBenchmark

I ran benchmarks of Qwen3.5 35b through 0.8b models using all different quants and compared performance with flash attention enabled vs disabled. Using llama.cpp and Unsloth’s UD_K_XL versions. Specifically for the Qwen…

TensorX

CoverageBenchmark

The BenchmarkList page for Qwen3.5-9B lists 101 benchmark rows and prices the model at $0.10 per 1M input tokens and $0.15 per 1M output tokens, situating it alongside comparison rows for Qwen Fable 5.1, Claude Fable 5.1, Claude Opus 5, Kimi K3, Qwen3.8-Flash-Next, Qwen3.8-2.4T-A95B, GLM 5.3, and GLM 5.3 Flash. In the The page distinguishes Qwen3.5-9B from sibling and successor variants by tagging the exact slug qwen-qwen3.5-9b and listing it next to Qwen3.8-Flash-Next and Qwen3.8-2.4T-A95B, allowing direct comparison against newer Qwen3.8 releases on the same benchmark rows. It functions as a model-focused benchmark hub rather than

Mixlayer

CoverageBenchmark

OpenRouter's first-party catalog page for qwen/qwen3.5-9b documents the model as a multimodal foundation model in the Qwen 3.5 family, using a unified vision-language design with early fusion of multimodal tokens so it can reason across text and images in a single context. The page confirms a 262,144-token context wind The listing also exposes routing modes (Balanced, Nitro, Exacto), a "Compare" playground, and an API key flow, giving developers concrete integration context for using qwen/qwen3.5-9B behind an OpenAI-compatible interface. Capabilities called out include reasoning, tool calling, structured output, attachments, open wei