Models·OpenAI

gpt-oss-20b: Open-Weight Model Pricing and Features

openai/gpt-oss-20b

An open-weight OpenAI text model with reasoning, tool support, and low per-token prices.

$0.018Input, per million tokens
$0.090Output, per million tokens
131KContext window
33KLongest answer

131K tokens of context — about 98,304 words, on a scale that ends at 2M.

What it can do

  • Reasoning
  • Tool use
  • Open weights
  • Takes text
  • Gives text

Released 5 Aug 2025 · checked 3 Oct 2026.

About it

gpt-oss-20b is an open-weight model from OpenAI, released under the Apache 2.0 license. It handles text input and output, supports reasoning and tools, and uses a mixture-of-experts design. It can read up to 131,072 tokens at once and produce up to 32,768 tokens in one answer.

Review

In a line

gpt-oss-20b is an open-weight language model made by OpenAI. It uses a mixture-of-experts design, with 21 billion parameters overall and 3.6 billion active for each pass.

What it's good at

The model supports reasoning and tool use, making it suited to text tasks that need step-by-step work or actions through connected tools.

It accepts and returns text only. It does not provide image input or image output. Its Apache 2.0 license and open weights give users access to the model files. See other models with tool calling or OpenAI models.

Price in plain terms

Input costs $0.018 per million tokens, and output costs $0.09 per million tokens. Tokens are small pieces of text counted for billing.

Reading a page of about 500 words costs under 0.01 cents. One chat turn with about 1,000 input tokens and 500 output tokens also costs under 0.01 cents. A thousand such turns cost about $0.06.

Context and limits

The context window is 131,072 tokens, so one request can include a large amount of text. A single answer can contain up to 32,768 tokens.

This is a text-only model. Its listed capabilities cover reasoning and tools, not images.

Ways to use it

Use gpt-oss-20b in an app or chat tool that offers the model and its tool support. Its model identifier is openai/gpt-oss-20b, which providers can use when exposing it through their services.

Verdict

Choose this model if you want OpenAI's open-weight option, need reasoning or tool use, and care about keeping token costs low. The Apache 2.0 license may also suit projects that need open model weights.

Skip it if your work needs images, because the model handles text only. You should also check that the app or provider you plan to use offers this model and its 131,072-token context window.

Verdict

gpt-oss-20b fits people comparing low-cost text models who want open weights, a permissive Apache 2.0 license, and support for reasoning and tools. Its large context window helps with long text, while the 32,768-token answer limit covers substantial replies. The model is not a match for image-based work, since both input and output are text. Availability depends on an app, chat service, or provider offering the model under its identifier.

  • Open weights under the Apache 2.0 license
  • Reasoning and tool support
  • 131,072-token context window
  • About $0.06 for 1,000 typical chat turns
  • Low input and output token prices
  • Text input and output only
  • Image tasks are not supported
  • Requires a service that offers this model
  • Pick it ifText tasks needing reasoning or tool use; Buyers comparing low per-token costs; Projects that want open model weights
  • Skip it ifYou need image input or image output; Your chosen app or provider does not offer gpt-oss-20b

Pricing

Input is $0.018 per million tokens and output is $0.09 per million tokens. A typical 1,000-chat-turn workload costs about $0.06.

Questions

What is gpt-oss-20b?

It is an open-weight, 21-billion-parameter text model from OpenAI, released under the Apache 2.0 license.

Can gpt-oss-20b reason or use tools?

Yes. The listed capabilities include reasoning and tool use.

How much does gpt-oss-20b cost?

Input costs $0.018 per million tokens and output costs $0.09 per million tokens. About 1,000 typical chat turns cost $0.06.

How much text can it handle?

Its context window is 131,072 tokens, and one answer can contain up to 32,768 tokens.

Does gpt-oss-20b work with images?

No. It accepts text and returns text only.

What is the model identifier?

The identifier is openai/gpt-oss-20b.

On these boards

More from OpenAI