Model · Alibaba · Text
Qwen3.5 9B / 4B
Alibaba's newest small open models — dense 9B (headline) and 4B under Apache 2.0, with image input, 262K context and thinking on by default. Qwen3.6 and Qwen3.8 ship nothing smaller than 27B, so this is still the current small Qwen.
- Modality
- Text
- License
- Apache 2.0 (Open)
- Parameter size
- 9B (dense); 4B sibling. 0.8B and 2B sizes also published
- Context window
- 262,144 tokens · up to 1.01M 262,144 tokens natively, extensible to about 1,010,000 with YaRN scaling. Alibaba advises keeping at least 128K so thinking works well.
- Released
- March 2, 2026
- Last verified
- October 28, 2026
- Runs locally
- Yes
- Also handles
- vision, reasoning, code
What you need to run it
- Mac
- Yes. Any Apple Silicon Mac (M1 or newer) with at least 8 GB (16 GB recommended) of memory. Newer chips are faster, but memory is what decides whether it runs.
- Windows / Linux PC
- NVIDIA works best. A graphics card with at least 6 GB (12 GB recommended) of video memory.
- No graphics card
- Works. Runs on a normal computer's processor, just slower.
The 4B runs on an 8 GB Mac and the 9B is comfortable on 16 GB, in Ollama, LM Studio or MLX.
Numbers are for: 4-bit builds: 4B (smallest) and 9B (headline, about 7 GB). Our estimate from the model’s size — the maker publishes no figure.
Strengths
- Apache 2.0 — no commercial restrictions
- Small enough for a laptop or a single consumer graphics card — about 8 GB of graphics memory for the 9B at 4-bit (our estimate, not an Alibaba figure)
- Native vision-language: reads images and documents as well as text
- Long 262K context for a model this size
- Thinking and non-thinking modes, plus tool calling
Weaknesses
- Thinking is on by default, which adds latency unless you switch it off
- Release date comes from press coverage — the model card carries none
- Older than Qwen3.6 and Qwen3.8 — the newer generations' gains are only available from 27B up
Try it
| Where | Type | Notes |
|---|---|---|
| Hugging Face (9B) | weights | Apache 2.0 |
| Hugging Face (4B) | weights | Apache 2.0 |
| Ollama | local | Quantised builds |
Official sources
- Model card (9B) model
- Model card (4B) model
Change log
- — Added hardware requirements (Mac support, memory, NVIDIA). Source: estimate from size; Ollama qwen3.5 page (9B 6.6 GB); mlx-community/Qwen3.5-9B-MLX-4bit.
- — Initial entry. Added 2026-10-28 from the popularity sweep: 6.18M Hugging Face downloads/month for Qwen3.5-9B and 7.41M for Qwen3.5-4B. Release date 2026-03-02 is from press coverage of the Qwen team's small-model launch — the model card gives no date and the Hugging Face repos were created 2026-02-27. Apache 2.0, 262,144-token context, free to self-host.
Esc