Qwen3.5 122B-A10B
Qwen3.5 122B-A10B is a mid-sized mixture-of-experts model from Alibaba with 10 billion active parameters. It understands images, handles 262,000 tokens of context and, quantised, fits a Mac Studio with 128 GB. That makes it a department server that can read long documents in one pass.
Profile
- Maker
- Alibaba (Qwen) · China
- Licence
- Free for commercial use
- Runs on
- Mac Studio with 128 GB
- Memory at 4 bits
- 88 GB
- Memory at 8 bits
- 130 GB
- Parameters
- 125.1bn, 10bn active
- Context
- 262,000 tokens
- Inputs
- text, image
- Released
- February 2026
Scores
- Human rating
- 1417 · rank 43 of 226 open
- In German
- 1425 · rank 27 of 154 open
- LMArena · Coding
- 1436 · 43/221
- LMArena · Maths
- 1431 · 31/215
- LMArena · Creative writing
- 1368 · 47/224
- LMArena · Following instructions
- 1399 · 44/226
- LMArena · Hard prompts
- 1421 · 45/226
- LMArena · Long prompts
- 1410 · 45/205
- LMArena · Multi-turn dialogue
- 1417 · 37/224
- LMArena · Building websites (WebDev)
- 1361 · 38/58
- LMArena · Understanding images (Vision)
- 1245 · 15/49
Scores: LMArena, Epoch AI, EuroEval, Hugging Face, models.dev; as of 19 October 2026. Sources and method →
Use cases
Where the model makes sense.
Long documents in one pass
- Situation
- Tenders, expert reports or project files run to hundreds of pages.
- What it looks like
- The model reads the documents in full and answers questions about them, such as deadlines, eligibility criteria or contradictions.
- What to watch
- Long contexts need a lot of memory and time. For recurring queries, document search (RAG) is more efficient.
Plans, scans and diagrams
- Situation
- In construction, engineering or purchasing, information arrives as plans, sketches or scans.
- What it looks like
- The model describes plans and scans, reads labels and tables and summarises them.
- What to watch
- Always check dimensions and quantities from plans in the original; the model can mismatch labels.
Knowledge assistant for a department
- Situation
- A department with many documents should find answers to internal questions faster.
- What it looks like
- AnythingLLM or LibreChat with document search; the model runs on a Mac Studio in the department.
- What to watch
- Carry over document access rights so nobody sees things through the AI that they could not see otherwise.
Translating technical documentation
- Situation
- Operating manuals and data sheets should be translated into several languages.
- What it looks like
- The model translates section by section with a glossary of technical terms, and the specialist department reviews.
- What to watch
- Safety notices in manuals are subject to their own requirements and need expert review.
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
If you only have a computer with 32 GB, take Qwen 3.8 27B. For a European maker in the same hardware class, Mistral Medium 3.5 fits.
How to start
ollama run qwen3.5:122bWith Ollama on a Mac Studio with 128 GB in the 4-bit version; at full context length memory gets tight.
Suitable interfaces and programs from our directory: Ollama, LibreChat, AnythingLLM
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.
OpenAI · United States
gpt-oss-120b
OpenAI's open text model from 2025 with adjustable reasoning effort and tool calls. Fits on a Mac Studio with 128 GB or a single 80 GB GPU.
Free for commercial use · Runs on: Mac Studio with 128 GB
Use cases and details →Mistral AI · France
Mistral Medium 3.5
Mid-sized Mistral AI model with image input and 262,000 tokens of context; quantised, it fits a Mac Studio with 128 GB. The licence excludes large companies.
Commercial use with conditions · Runs on: Mac Studio with 128 GB
Use cases and details →Alibaba (Qwen) · China
Qwen 3.8 27B
Compact model from the Qwen 3.8 series that understands images and calls tools. Quantised, it runs on a computer with 32 GB and reaches scores otherwise seen from much larger models.
Free for commercial use · Runs on: Computer with 32 GB
Use cases and details →
FAQ
What hardware do I need for Qwen3.5 122B-A10B?
In the common 4-bit version, Qwen3.5 122B-A10B needs about 88 GB of memory including headroom for context and runtime. That fits: Mac Studio with 128 GB. The more accurate 8-bit version needs about 130 GB. Long documents increase the requirement.
May I use Qwen3.5 122B-A10B commercially?
The licence allows commercial use, modification and redistribution; the main obligation is to keep the licence notice.
How good is Qwen3.5 122B-A10B in German?
On LMArena, people rate Qwen3.5 122B-A10B at 1425 points for prompts in German, rank 27 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.