Antimatter
OpenAI

gpt-oss-20b

openai/gpt-oss-20b

gpt-oss-20b is an open-weight 21B parameter model released by OpenAI under the Apache 2.0 license.

$0.033 in · $0.143 out per 1M tokens131k contextAug 5, 2025

less than 1/100 of a cent a message

A typical message: about 500 tokens you send and 500 it writes.

Also:batch

Published facts from the catalog, billed prices and what we measured here. Anything not measured says so.

Price, billed here

Input, per 1M tokens
$0.033provider $0.03
Output, per 1M tokens
$0.143provider $0.13
Cached input, per 1M
$0.033provider $0.03
Margin
+10%

Context and formats

Context window
131,072 tokens
Longest answer
117,964 tokens
Reads
text
Makes
text
First listed
2025-08-05

Measured here, last 7 days

First token, median
not measured
Answered without error
not measured

Tools and JSON

Tool calling
declaredthe provider's claim, not yet measured here
JSON output
declaredthe provider's claim, not yet measured here

Price and latency history as JSON: GET /api/planet/history?id=openai/gpt-oss-20b.

gpt-oss-20b is an open-weight 21B parameter model released by OpenAI under the Apache 2.0 license. It uses a Mixture-of-Experts (MoE) architecture with 3.6B active parameters per forward pass, optimized for...

In OpenAI's words.

The answer appears here.

Use one OpenAI compatible endpoint, this model's id and an am_live_ key from the dashboard. You can also try it in chat first with no key.

curl https://useantimatter.com/v1/chat/completions \
  -H "Authorization: Bearer $ANTIMATTER_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"model": "openai/gpt-oss-20b", "messages": [{"role": "user", "content": "Hello"}]}'

Everything else is in the docs.

A badge showing this model's live price per million tokens, as billed here, for a README or a site. It links back to this page.

gpt-oss-20b price on Antimatter
<a href="https://useantimatter.com/planet/openai/gpt-oss-20b"><img src="https://useantimatter.com/api/badge/price?id=openai%2Fgpt-oss-20b" alt="gpt-oss-20b price on Antimatter" height="28"></a>

Closest in price and memory, from other labs.

Go back to the map, or take this model into a chat.