Explore additional models
Frequently
asked questions
GPT-OSS 20B is the smaller of OpenAI’s two open-weight (“OSS”) models, with 21B total parameters and about 3.6B active per token in a mixture-of-experts design. It carries the same reasoning training, response format, and reasoning-effort controls as GPT-OSS 120B, and ships under the Apache 2.0 license. On Crusoe it is available for Serverless Fine-Tuning and Self-Serve Deployments, with a 131,072-token context. You can also request a Tailored Deployment for bespoke inference optimization and a dedicated endpoint.
GPT-OSS 120B is the higher-quality model, while GPT-OSS 20B is faster, smaller, and cheaper to serve and tune. Both share the same training, response format, and low, medium, and high reasoning-effort controls. On Crusoe, 120B is available for Serverless Fine-Tuning, Serverless Inference, and Self-Serve Deployments. The 20B model is available for Serverless Fine-Tuning and Self-Serve Deployments, where its single-GPU footprint is the advantage.
GPT-OSS 20B is billed by how you use it. Serverless Fine-Tuning is billed per 1M training tokens, where training tokens equal dataset tokens multiplied by epochs. Self-Serve Deployments are billed per GPU hour based on the GPU you choose, giving you predictable costs at production volume. Current rates are shown in the pricing module above. Log in or create a free account to get started quickly in Crusoe Intelligence Foundry.
OpenAI reports GPT-OSS 20B performing comparably to o3-mini on common reasoning benchmarks while running in about 16GB of memory, which makes it one of the strongest small models available for a dedicated endpoint. Like 120B, it supports native tool use, structured output, and chain-of-thought reasoning.
Yes. Use Serverless Fine-Tuning to customize GPT-OSS 20B on your own data. Upload a JSONL or Parquet dataset, run a LoRA job with pre-configured defaults, then go live in one click using Self-Serve Deployments or download the weights to deploy anywhere. Self-Serve Deployments offer inference profiles optimized for throughput or responsiveness. You can also request a Tailored Deployment for bespoke inference optimization and a dedicated endpoint.
Yes, GPT-OSS 20B supports three reasoning-effort levels: low, medium, and high, set in the system prompt, with medium as the default. Higher effort produces longer internal reasoning and better results on hard problems, at the cost of latency and output tokens. On a dedicated endpoint, lower effort on simple requests frees up capacity, so the same GPU can serve more traffic.
No, running GPT-OSS 20B on Crusoe never sends prompts, outputs, or fine-tuning data to OpenAI. GPT-OSS is an open-weight model, so Crusoe serves it on Crusoe-operated infrastructure rather than routing requests through OpenAI’s API. Data handling is governed by Crusoe’s infrastructure and policies, not by OpenAI.


%204.png)
