Qwen 3.8 27B
Qwen 3.8 27B from Alibaba is a compact model that understands images, calls tools and can reason before answering. Quantised, it runs on a computer with 32 GB and reaches scores in independent tests otherwise seen from much larger models. The Apache 2.0 licence allows commercial use without conditions.
Profile
- Maker
- Alibaba (Qwen) · China
- Licence
- Free for commercial use
- Runs on
- Computer with 32 GB
- Memory at 4 bits
- 20 GB
- Memory at 8 bits
- 29 GB
- Parameters
- 27.8bn
- Context
- 262,000 tokens
- Inputs
- text, image
- Released
- August 2026
Scores
- Human rating
- 1441 · rank 19 of 226 open
- In German
- 1436 · rank 22 of 154 open
- Epoch index
- 151.6 · rank 7 of 127 open
- LMArena · Coding
- 1483 · 12/221
- LMArena · Maths
- 1457 · 17/215
- LMArena · Creative writing
- 1384 · 39/224
- LMArena · Following instructions
- 1437 · 13/226
- LMArena · Hard prompts
- 1459 · 12/226
- LMArena · Long prompts
- 1449 · 13/205
- LMArena · Multi-turn dialogue
- 1442 · 22/224
- LMArena · Building websites (WebDev)
- 1592 · 11/58
- LMArena · Understanding images (Vision)
- 1271 · 5/49
Scores: LMArena, Epoch AI, EuroEval, Hugging Face, models.dev; as of 19 October 2026. Sources and method →
Use cases
Where the model makes sense.
Coding assistant on your own network
- Situation
- A development team wants AI help with coding but may not send client source code to cloud services.
- What it looks like
- The model runs on a machine on the network, and a coding agent such as Pi or an editor extension connects to it. Code, tests and error messages stay in-house.
- What to watch
- Run agents with write access in an isolated environment and review changes before accepting them.
Agents for recurring processes
- Situation
- Requests from forms and mailboxes have to be read, classified and routed to the right place.
- What it looks like
- Using tool calls (for example via MCP), the model reads the request, looks up customer data in the CRM and creates a ticket with category and summary.
- What to watch
- Start with read access and allow write actions only after a test phase.
Analysing technical documents
- Situation
- Data sheets, test reports and standards need comparing, for example when choosing components.
- What it looks like
- The model reads several documents with tables and compares key values; with 262,000 tokens of context, longer documents fit into one request.
- What to watch
- Spot-check values from tables against the original, especially units.
Understanding screenshots and charts
- Situation
- Error messages, dashboards and charts reach support as images.
- What it looks like
- The model describes what the screenshot shows, reads values from charts and suggests possible causes.
- What to watch
- For unlabelled charts the model estimates values; the answer should make that clear.
The use cases were written with AI assistance (Claude by Anthropic) and reviewed by vona. Scores come unchanged from the sources listed.
When another model fits better
Human raters score it lower for creative writing than for coding and analysis. For marketing copy and correspondence in German, Gemma 4 31B or Mistral Small 3.2 in the same hardware class are a better fit.
How to start
ollama run qwen3.8:27bWith Ollama on a computer with 32 GB. Reasoning mode pays off for agents and coding; for short answers the normal mode is enough.
Suitable interfaces and programs from our directory: Ollama, LibreChat, Pi
Licence: Free for commercial use
The licence allows commercial use, modification and redistribution; the main obligation is to keep the licence notice.
Hugging Face ↗Qwen as a cloud service in the AI tools directory →Similar models
Comparable models.
Google · United States
Gemma 4 31B
Compact Google model with image input under the Apache 2.0 licence. It ranks near the top of EuroEval's German comparison and runs on a computer with 32 GB.
Free for commercial use · Runs on: Computer with 32 GB
Use cases and details →Alibaba (Qwen) · China
Qwen 3.6 35B-A3B
Qwen mixture-of-experts model with only three billion active parameters per request, so it responds quickly even on ordinary hardware; image input included.
Free for commercial use · Runs on: Workstation with 64 GB
Use cases and details →OpenAI · United States
gpt-oss-20b
The small version of gpt-oss with adjustable reasoning effort and tool calls, for computers with 32 GB.
Free for commercial use · Runs on: Computer with 32 GB
Use cases and details →
FAQ
What hardware do I need for Qwen 3.8 27B?
In the common 4-bit version, Qwen 3.8 27B needs about 20 GB of memory including headroom for context and runtime. That fits: Computer with 32 GB. The more accurate 8-bit version needs about 29 GB. Long documents increase the requirement.
May I use Qwen 3.8 27B commercially?
The licence allows commercial use, modification and redistribution; the main obligation is to keep the licence notice.
How good is Qwen 3.8 27B in German?
On LMArena, people rate Qwen 3.8 27B at 1436 points for prompts in German, rank 22 of 154 open models. The best way to see whether it suits your texts is a test with your own examples.
AI consulting
Your own language model for your company?
Together we find the model that fits your tasks, your hardware and your data protection requirements, test it with real documents and set up operation.