GPT-OSS 20B (NVIDIA) API

Published API pricing

Rates below are in USD per 1 million tokens. Missing rates are not free usage. Check media-specific billing before using a token-only estimate for an image or video workflow.

Cache rates apply only when supported and reported for the request. A lower cache rate does not guarantee a cache hit. Check this model in the live price comparison.

Context and output limits

Published context window

131,072 tokens

Maximum output

8,192 tokens

Request overhead and output need to fit the selected endpoint’s limits. Model publication does not guarantee uninterrupted availability or every optional feature.

Connect GPT-OSS 20B (NVIDIA) to your application

1. Review access options and activate the access required by the model. Check pay-as-you-go eligibility for this model.
2. Create a key in your account’s API Keys area. Keep it server-side.
3. Choose a supported format in the API documentation, and use the exact model ID shown above.
4. Test a small request before adding streaming, tools or large prompts.

OpenAI-compatible base URL: https://api.g0i.ai/v1. Native Anthropic clients use https://api.g0i.ai. Endpoint support varies by model; these are gateway addresses, not a promise that both formats support every model feature.

Questions before you integrate

Does the published rate include subscription access?

No. Per-token prices and subscription eligibility are separate. Review the current access plans and account details.

Can I swap another model ID into the same request?

Only after checking endpoint compatibility and request fields. Validate tool schemas, output parsing and context limits with the model you choose.

What if a rate or limit is missing?

Do not assume zero cost or unlimited capacity. Contact support about requirements that are important to your application.

Compare other published models

• Gemini 2.5 Flash Image (Vision)
• Gemini 2.5 Flash
• GPT-OSS 120B
• MiniMax M2.5
• GPT-5.4
• GPT-5.1