GPT OSS 120B

• Type: Generation
• Capabilities: reasoning
• Cache read: $0.11 per 1M input tokens (0.75× standard input price). See prompt caching.

Overview

Meet gpt-oss-20b — OpenAI’s larger open-weight MoE model, with 117B total parameters, 5.1B active parameters, and a 128k-token context window.

Reasoning efforts

• Supported: low, medium, high

See the reasoning effort guide for request examples.

Pricing

Priority	Input Tokens (per 1M)	Cache Read Tokens (per 1M)	Output Tokens (per 1M)
Realtime1	$0.15	$0.11	$0.60
Async	$0.11	$0.08	$0.45
Batch (24h)	$0.08	$0.06	$0.30

Playground

Open this model in the Playground.

Footnotes

1. Realtime availability is limited. Doubleword is primarily a batch API. ↩