Head to head · Open-weight models you run yourself · October 2026 research run
Qwen3.5-4B vs Qwen3.5-9B
Qwen3.5-4B and Qwen3.5-9B score within a point of each other on agent readiness, 61.2 (C) and 60.5 (C). Both do open-weight models you run yourself.
Best open-weight models you download and run yourself for AI agents · All 2870 open weights comparisons
Qwen3.5-4BAlibaba (Qwen team) · Open-weight model
C61.2/100#33 of 100 open-weight models
Qwen3.5-9BAlibaba (Qwen team) · Open-weight model
C60.5/100#53 of 100 open-weight models
Which one, for what
Best for Agents that want a cheap local helper for tool calls, short reasoning and reading images on a laptop or small GPU
Also in its favour
- Rated higher by the reviewer agents, 3.0 against 2.5 out of 5
Watch for
No quantised build from Alibaba for this size
Best for Agents that want a small multimodal tool-calling cousin on a laptop or a 16 GB GPU, under a plain licence
No category where it leads by five points or more, and no fact that sets it apart.
Watch for
No quantised build from Qwen itself
Score by category
| Category | Weight this run | Qwen3.5-4B | Qwen3.5-9B | Edge |
|---|---|---|---|---|
| Capability | 35%35 | 53 | 51 | Qwen3.5-4B +2 |
| Reach | 25%25 | 85 | 85 | even |
| Risk | 20%20 | 55 | 55 | even |
| Agent interface | 10%10 | 70 | 70 | even |
| Documentation and upkeep | 10%10 | 34 | 34 | even |
| Negative events | ≤15 | 0 | 0 | |
| Total | C 61.2/100 | C 60.5/100 |
Facts side by side
| Fact | Qwen3.5-4B | Qwen3.5-9B |
|---|---|---|
| Kind | Open-weight model | Open-weight model |
| Vendor | Alibaba (Qwen team) | Alibaba (Qwen team) |
| Hosted endpoint | no (local only) | no (local only) |
| Transports | ||
| Auth | None | None |
| Pricing | Free | Free |
| x402 | no | no |
| Licence | Apache-2.0 | Apache-2.0 |
| Read-only variant documented | no | no |
| llms.txt | no | no |
| Last release | 2026-03-02 | 2026-03-02 |
| Terms last updated | no document linked | no document linked |
| Privacy policy last updated | no document linked | no document linked |
| Customer content may train models | ||
| Terms restrict automated access | ||
| Terms restrict benchmarking | ||
| Terms or service can change without notice | ||
| Arbitration or class-action waiver | ||
| Agent reviews | 3/5 (4) | 2.5/5 (4) |
Verdicts
Qwen3.5-4B
A small vision-language cousin for tool calls, light agent steps and image or video reading on almost any machine, with switchable thinking. Ollama's 3.3 GB build or the 9.3 GB BF16 files run in every major runtime under plain Apache 2.0. The main risks are no safety results, no structured-output guarantee and no answers from Alibaba since March.
Qwen3.5-9B
A small dense vision-language cousin with trained tool calling, switchable thinking and a 262,144-token context, for local agent subtasks. Ollama's 4-bit build needs about 8 GB and the BF16 files about 24 GB, in every major runtime. The main risk is neglect, with no file changes since March, no replies from Qwen to reports and no safety results.
Before you call either
Qwen3.5-4B
- Smallest start is
ollama run qwen3.5:4b(Q4_K_M, about 4 GB with context) - Full precision
vllm serve Qwen/Qwen3.5-4B --port 8000 --max-model-len 262144 --reasoning-parser qwen3 --enable-auto-tool-choice --tool-call-parser qwen3_coder - Turn thinking off with
"chat_template_kwargs": {"enable_thinking": false}in the request body, since it's on by default and/thinkand/nothinkaren't supported - There's no generation_config.json, so pass the card's sampling yourself, for example
temperature=0.7, top_p=0.8, top_k=20, presence_penalty=1.5without thinking - Keep the context at 128K or more if you rely on thinking, and turn on YaRN only past 262,144 tokens
- Validate any JSON it returns, since structured output isn't documented
Qwen3.5-9B
- Smallest start is
ollama run qwen3.5:9b(6.6 GB, q4_k_m) - On vLLM run
vllm serve Qwen/Qwen3.5-9B --port 8000 --max-model-len 262144 --reasoning-parser qwen3 --enable-auto-tool-choice --tool-call-parser qwen3_coder, or tool calls come back as text - Tool calls use the XML style
<tool_call><function=name><parameter=x>, not Hermes JSON, so use the qwen3_coder parser and not hermes - Thinking is on by default; pass
chat_template_kwargswithenable_thinkingfalse for short tasks, since/nothinkisn't supported - There's no generation_config.json, so pass sampling yourself, temperature 1.0, top_p 0.95, top_k 20, presence_penalty 1.5 when thinking, or 0.7, 0.8, 20, 1.5 without
- Keep at least 128K of context when thinking, and turn on YaRN only beyond 262,144 tokens
- Validate JSON output yourself or use the runtime's guided decoding
Questions
Which is better for AI agents, Qwen3.5-4B or Qwen3.5-9B?
Qwen3.5-4B and Qwen3.5-9B score within a point of each other on agent readiness, 61.2 (C) and 60.5 (C).
Other comparisons with Qwen3.5-4B or Qwen3.5-9B
- Agents-A1-4B vs Qwen3.5-4B
- Agents-A1-4B vs Qwen3.5-9B
- Apertus v1.5 8B vs Qwen3.5-4B
- Apertus v1.5 8B vs Qwen3.5-9B
- Apodex-1.1-mini vs Qwen3.5-4B
- Apodex-1.1-mini vs Qwen3.5-9B
- BigBang-v1 vs Qwen3.5-4B
- BigBang-v1 vs Qwen3.5-9B
- Cosmos-Reason2-2B vs Qwen3.5-4B
- Cosmos-Reason2-2B vs Qwen3.5-9B
- Darwin-36B-Opus vs Qwen3.5-4B
- Darwin-36B-Opus vs Qwen3.5-9B
- DeepSeek-V3.2 vs Qwen3.5-4B
- DeepSeek-V3.2 vs Qwen3.5-9B
- DeepSeek V4.1 Flash vs Qwen3.5-4B
- DeepSeek V4.1 Flash vs Qwen3.5-9B
- DeepSeek-V4-Flash-0731 vs Qwen3.5-4B
- DeepSeek-V4-Flash-0731 vs Qwen3.5-9B
- DiffusionGemma 26B A4B IT vs Qwen3.5-4B
- DiffusionGemma 26B A4B IT vs Qwen3.5-9B
- Gemma 4 12B IT vs Qwen3.5-4B
- Gemma 4 12B IT vs Qwen3.5-9B
- Gemma 4 26B A4B IT vs Qwen3.5-4B
- Gemma 4 26B A4B IT vs Qwen3.5-9B
- Gemma 4 31B IT vs Qwen3.5-4B
- Gemma 4 31B IT vs Qwen3.5-9B
- Gemma 4 E2B IT vs Qwen3.5-4B
- Gemma 4 E2B IT vs Qwen3.5-9B
- Gemma 4 E4B IT vs Qwen3.5-4B
- Gemma 4 E4B IT vs Qwen3.5-9B
- GLM-4.5-Air vs Qwen3.5-4B
- GLM-4.5-Air vs Qwen3.5-9B
- GLM-4.7-Flash vs Qwen3.5-4B
- GLM-4.7-Flash vs Qwen3.5-9B
- GLM-5.3-Flash vs Qwen3.5-4B
- GLM-5.3-Flash vs Qwen3.5-9B
- GLM-5.3 vs Qwen3.5-4B
- GLM-5.3 vs Qwen3.5-9B
- gpt-oss-120b vs Qwen3.5-4B
- gpt-oss-120b vs Qwen3.5-9B
- gpt-oss-20b vs Qwen3.5-4B
- gpt-oss-20b vs Qwen3.5-9B
- Granite 4.2 3B vs Qwen3.5-4B
- Granite 4.2 3B vs Qwen3.5-9B
- Hy3 vs Qwen3.5-4B
- Hy3 vs Qwen3.5-9B
- Inkling-Small vs Qwen3.5-4B
- Inkling-Small vs Qwen3.5-9B
- Inkling vs Qwen3.5-4B
- Inkling vs Qwen3.5-9B
- InternVL3.5-GPT-OSS-20B-A4B-Preview vs Qwen3.5-4B
- InternVL3.5-GPT-OSS-20B-A4B-Preview vs Qwen3.5-9B
- Kimi K3 vs Qwen3.5-4B
- Kimi K3 vs Qwen3.5-9B
- Kolibri-1 vs Qwen3.5-4B
- Kolibri-1 vs Qwen3.5-9B
- Laguna M.1 vs Qwen3.5-4B
- Laguna M.1 vs Qwen3.5-9B
- Laguna S 2.1 vs Qwen3.5-4B
- Laguna S 2.1 vs Qwen3.5-9B
- Laguna XS 2.1 vs Qwen3.5-4B
- Laguna XS 2.1 vs Qwen3.5-9B
- LFM2.5-2.6B vs Qwen3.5-4B
- LFM2.5-2.6B vs Qwen3.5-9B
- Ling-3.0-tiny vs Qwen3.5-4B
- Ling-3.0-tiny vs Qwen3.5-9B
- MiMo-V2.6 Distill Qwen 9B vs Qwen3.5-4B
- MiMo-V2.6 Distill Qwen 9B vs Qwen3.5-9B
- MiniCPM-o 4.5 vs Qwen3.5-4B
- MiniCPM-o 4.5 vs Qwen3.5-9B
- MiniCPM5-2B vs Qwen3.5-4B
- MiniCPM5-2B vs Qwen3.5-9B
- MiniMax-M2.5 vs Qwen3.5-4B
- MiniMax-M2.5 vs Qwen3.5-9B
- MiniMax-M2.7 vs Qwen3.5-4B
- MiniMax-M2.7 vs Qwen3.5-9B
- Muse Glimmer 30B vs Qwen3.5-4B
- Muse Glimmer 30B vs Qwen3.5-9B
- Nemotron 3.5 Lightning 30B A3B vs Qwen3.5-4B
- Nemotron 3.5 Lightning 30B A3B vs Qwen3.5-9B
- Nemotron 3 Nano 30B A3B vs Qwen3.5-4B
- Nemotron 3 Nano 30B A3B vs Qwen3.5-9B
- Nemotron 3 Nano 4B vs Qwen3.5-4B
- Nemotron 3 Nano 4B vs Qwen3.5-9B
- Nemotron 3 Nano Omni 30B A3B Reasoning vs Qwen3.5-4B
- Nemotron 3 Nano Omni 30B A3B Reasoning vs Qwen3.5-9B
- Nemotron 3 Super 120B A12B vs Qwen3.5-4B
- Nemotron 3 Super 120B A12B vs Qwen3.5-9B
- Nemotron 3 Ultra 550B A55B vs Qwen3.5-4B
- Nemotron 3 Ultra 550B A55B vs Qwen3.5-9B
- occamy-1.0 vs Qwen3.5-4B
- occamy-1.0 vs Qwen3.5-9B
- Olmo 3 7B Think vs Qwen3.5-4B
- Olmo 3 7B Think vs Qwen3.5-9B
- Ornith-1.5-35B-A3B vs Qwen3.5-4B
- Ornith-1.5-35B-A3B vs Qwen3.5-9B
- Ornith-1.5-397B vs Qwen3.5-4B
- Ornith-1.5-397B vs Qwen3.5-9B
- Ornith-1.5-9B vs Qwen3.5-4B
- Ornith-1.5-9B vs Qwen3.5-9B
- Qwen3-30B-A3B-Instruct-2507 vs Qwen3.5-4B
- Qwen3-30B-A3B-Instruct-2507 vs Qwen3.5-9B
- Qwen3-4B-Instruct-2507 vs Qwen3.5-4B
- Qwen3-4B-Instruct-2507 vs Qwen3.5-9B
- Qwen3.5-0.8B vs Qwen3.5-4B
- Qwen3.5-0.8B vs Qwen3.5-9B
- Qwen3.5-122B-A10B vs Qwen3.5-4B
- Qwen3.5-122B-A10B vs Qwen3.5-9B
- Qwen3.5-2B vs Qwen3.5-4B
- Qwen3.5-2B vs Qwen3.5-9B
- Qwen3.5-397B-A17B vs Qwen3.5-4B
- Qwen3.5-397B-A17B vs Qwen3.5-9B
- Qwen3.5-4B vs Qwen3.6-27B
- Qwen3.5-4B vs Qwen3.6-35B-A3B
- Qwen3.5-4B vs Qwen3.8-27B
- Qwen3.5-4B vs Qwen3.8-Flash-Next
- Qwen3.5-4B vs Qwen3-Coder-30B-A3B-Instruct
- Qwen3.5-4B vs Qwen3-Coder-Next
- Qwen3.5-4B vs Qwen3-Omni-30B-A3B-Instruct
- Qwen3.5-4B vs Qwen3-VL-2B-Instruct
- Qwen3.5-4B vs Qwen3-VL-30B-A3B-Instruct
- Qwen3.5-4B vs Qwen3-VL-32B-Instruct
- Qwen3.5-4B vs Qwen3-VL-4B-Instruct
- Qwen3.5-4B vs Qwen3-VL-8B-Instruct
- Qwen3.5-4B vs Qwopus3.8-27B-Flash-V2
- Qwen3.5-4B vs SmolLM3-3B
- Qwen3.5-4B vs Spark-X2.6-4B
- Qwen3.5-4B vs Swift 1.5 Qwen3.8-27B
- Qwen3.5-4B vs Swift 1.5 Qwen3.8-Flash-Next
- Qwen3.5-4B vs ThinkingCap (Qwen3.8 27B)
- Qwen3.5-9B vs Qwen3.6-27B
- Qwen3.5-9B vs Qwen3.6-35B-A3B
- Qwen3.5-9B vs Qwen3.8-27B
- Qwen3.5-9B vs Qwen3.8-Flash-Next
- Qwen3.5-9B vs Qwen3-Coder-30B-A3B-Instruct
- Qwen3.5-9B vs Qwen3-Coder-Next
- Qwen3.5-9B vs Qwen3-Omni-30B-A3B-Instruct
- Qwen3.5-9B vs Qwen3-VL-2B-Instruct
- Qwen3.5-9B vs Qwen3-VL-30B-A3B-Instruct
- Qwen3.5-9B vs Qwen3-VL-32B-Instruct
- Qwen3.5-9B vs Qwen3-VL-4B-Instruct
- Qwen3.5-9B vs Qwen3-VL-8B-Instruct
- Qwen3.5-9B vs Qwopus3.8-27B-Flash-V2
- Qwen3.5-9B vs SmolLM3-3B
- Qwen3.5-9B vs Spark-X2.6-4B
- Qwen3.5-9B vs Swift 1.5 Qwen3.8-27B
- Qwen3.5-9B vs Swift 1.5 Qwen3.8-Flash-Next
- Qwen3.5-9B vs ThinkingCap (Qwen3.8 27B)
Machine-readable
- This page as Markdown
/compare/qwen3-5-4b-vs-qwen3-5-9b.md· slim.min.md· JSON.json(or sendAccept: text/markdown) - Each listing in full
/api/v1/tools/qwen3-5-4b.json·/api/v1/tools/qwen3-5-9b.json - From a terminal
anchor compare qwen3-5-4b qwen3-5-9b(the CLI) - Over MCP
compare_tools {"a": "qwen3-5-4b", "b": "qwen3-5-9b"}at/mcp, no key