gpt-oss-20b
OpenAI · United States · Apache 2.0
Commercial use: Yes — free for commercial use
Apache 2.0, no conditions.
OpenAI's own model, published under Apache 2.0, small enough for a 16 GB machine, with a built-in dial for how hard it thinks.
Why it should matter to you
Useful as a hedge. Keeping an open-weight model that behaves similarly to your hosted one means a price change or a deprecation notice is an inconvenience rather than a re-architecture. The reasoning-effort control is the practical part: cheap mode for routing and classification, expensive mode only where the task needs it.
How it connects to our work
We build that fallback path into agent systems as a matter of course. Not because the hosted model is worse, but because you cannot pin a model you do not hold.
From our field notesVendor lock-in is a hidden risk, not only a legacy-systems problem.
Native MXFP4; roughly 12-14 GB in practice.
Advantages
- Runs on a 16 GB machine while keeping the reasoning-effort control
- Apache 2.0 from a US lab
- Fast: 3.6B active parameters per token
- A safety-classifier sibling (gpt-oss-safeguard-20b) exists under the same licence for filtering pipelines
Disadvantages
- Same conservative refusal behaviour as the 120b
- Limited world knowledge — pair with retrieval rather than relying on recall
Reach for it when
Local agent work and task routing on modest hardware. A sensible default fallback model to keep alongside a hosted frontier one.
Where it falls down
Anything relying on the model already knowing your domain. Without retrieval it guesses confidently.
Running it
16 GB GPU or Mac. A strong default for local agents. See the hardware sizing tables for how that maps to specific chips and cards, and the quantisation guide for what you give up at each bit width.
Free tiers carrying this model
| Provider | Type | The catch | Live limits |
|---|---|---|---|
| Groq | Inference host | Free-plan limits are per-model and vary widely across the catalogue. Token-per-day caps bite before request caps on long prompts. | check → |
| OpenRouter | Aggregator | Free-pool membership changes without notice — a model you built on can stop being free, and the list above will drift. Data passes through OpenRouter and then the upstream provider, so check the privacy and routing settings before sending anything sensitive. | check → |
| Cerebras Inference | Inference host | Small model catalogue compared with the aggregators. You are choosing it for speed, not selection. | check → |
| NVIDIA NIM | Inference host | It is a credit grant, not a permanent free tier. When the credits run out you are on a paid plan or self-hosting the container. | check → |
| Cloudflare Workers AI | Edge inference | The Neuron accounting makes it hard to predict how many actual requests you get, because different models consume at very different rates. | check → |
Jurisdiction: United States
Best raw capability and the deepest tooling ecosystem. For EU personal data you are relying on a transfer framework rather than on data never leaving the bloc, so check whether your DPA and your customers accept that.
Watch for: Enterprise API tiers usually promise no training on your data; consumer tiers and free tiers frequently do not. The free tier is where this bites.
Frequently asked
Can I use gpt-oss-20b commercially?
Apache 2.0, no conditions.
What hardware do I need to run gpt-oss-20b?
16 GB GPU or Mac. A strong default for local agents. Weights alone are roughly 13 GB at Q4_K_M and 13 GB at Q8_0. Native MXFP4; roughly 12-14 GB in practice.. Add KV cache on top of that, which grows with your context length.
What licence is gpt-oss-20b released under?
Apache 2.0. Full commercial use, modification and redistribution. Patent grant included. The most permissive licence in common use for open-weight models.
Where can I use gpt-oss-20b for free?
Free tiers carrying it include Groq, OpenRouter, Cerebras Inference, NVIDIA NIM, Cloudflare Workers AI. Limits differ per provider and change often, so check each provider's own limits page. You can also self-host the weights, which has no rate limit at all.
Similar models
gpt-oss-120b
80 GB GPU, or 96 GB+ unified memory.
DeepSeek-V4-Flash
Hosted, or 256 GB+ unified memory at Q4. Not a laptop model.
GLM-5.3-Flash
80-96 GB at Q4_K_M. A workstation or Ultra-class Mac.
Qwen3.8 27B
24 GB GPU at Q4_K_M, or a 32 GB Mac at Q6_K.
Wiring gpt-oss-20b into something real?
We build the evaluation harness, the failover and the cost ceilings around a model like this, so it survives contact with production.
Maps to AI agent development, AI automation development and AI strategy and consulting. Or see it working: our case studies.