gpt-oss-120b
gpt-oss-120b is OpenAI’s larger open model. With only about 5 billion active parameters it answers quickly but needs about 70 GB of memory. That fits a Mac Studio with 128 GB or a single 80 GB GPU and is enough for a central AI server for a whole team.
Profile
- Maker
- OpenAI · United States
- Licence
- Free for commercial use
- Runs on
- Mac Studio with 128 GB
- Memory at 4 bits
- 72 GB
- Memory at 8 bits
- 63 GB
- Parameters
- 116.8bn, 5.1bn active
- Context
- 131,000 tokens
- Inputs
- text
- Released
- August 2025
Scores
- Human rating
- 1366 · rank 72 of 226 open
- In German
- 1354 · rank 67 of 154 open
- Epoch index
- 140 · rank 35 of 127 open
- GPQA Diamond (expert questions)
- 67.7 %
- AIME (maths)
- 88.9 %
- LMArena · Coding
- 1379 · 75/221
- LMArena · Maths
- 1389 · 67/215
- LMArena · Creative writing
- 1275 · 96/224
- LMArena · Following instructions
- 1318 · 83/226
- LMArena · Hard prompts
- 1363 · 73/226
- LMArena · Long prompts
- 1319 · 89/205
- LMArena · Multi-turn dialogue
- 1340 · 82/224
Scores: LMArena, Epoch AI, EuroEval, Hugging Face, models.dev; as of 19 October 2026. Sources and method →
Use cases
Where the model makes sense.
Central AI chat for the company
- Situation
- Staff use private accounts with ChatGPT and others, and nobody knows what data ends up there.
- What it looks like
- A Mac Studio or server on site runs the model, and LibreChat provides the interface with login via the company account.
- What to watch
- Write a short usage policy (what may go in, what may not) and train staff; the EU AI Act requires this as well.
Comparing contracts and quotes
- Situation
- Several quotes or draft contracts should be compared and checked for deviations.
- What it looks like
- The model builds a table with prices, terms and clauses and flags what deviates from the standard.
- What to watch
- Legal assessment remains a job for experts. The model shows differences; it does not assess them bindingly.
Following calculations
- Situation
- Costings, quantity surveys or technical calculations should be checked and explained.
- What it looks like
- With high reasoning effort the model walks through the calculation step by step and names assumptions.
- What to watch
- Double-check the final figures with a calculation tool.
Agents for internal processes
- Situation
- Recurring processes such as leave requests, orders or reports take time.
- What it looks like
- Using tool calls, the model drives the systems involved and creates drafts for approval.
- What to watch
- Start with one process, measure what it saves, and only then connect others.
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
It does not understand images, and human raters score other models higher on German prompts. For a server with image understanding, Qwen3.5 122B-A10B fits; for German, Mistral Medium 3.5.
How to start
ollama run gpt-oss:120bWith Ollama on a Mac Studio with 128 GB or an 80 GB GPU; the official release is already quantised to 4 bits (MXFP4).
Suitable interfaces and programs from our directory: Ollama, LibreChat, Open WebUI
Licence: Free for commercial use
The licence allows commercial use, modification and redistribution; the main obligation is to keep the licence notice.
Hugging Face ↗Similar models
Comparable models.
Alibaba (Qwen) · China
Qwen3.5 122B-A10B
Mid-sized Qwen model with image input and 262,000 tokens of context. With 4-bit quantisation it fits on a Mac Studio with 128 GB.
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 →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 gpt-oss-120b?
In the common 4-bit version, gpt-oss-120b needs about 72 GB of memory including headroom for context and runtime. That fits: Mac Studio with 128 GB. The more accurate 8-bit version needs about 63 GB. Long documents increase the requirement.
May I use gpt-oss-120b commercially?
The licence allows commercial use, modification and redistribution; the main obligation is to keep the licence notice.
How good is gpt-oss-120b in German?
On LMArena, people rate gpt-oss-120b at 1354 points for prompts in German, rank 67 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.