LLM Reference

GPT-OSS Models by OpenAI

OpenAIApache 2.0Open source
4 models2025Up to 131k ctxFrom $0.03/1M input

Details

ResearcherOpenAI
LicenseApache 2.0OSI-approved
Commercial useCommercial use: permitted
Models4
Released2025
Max context131k

Capabilities

Function Calling3 of 4 models
Tool Use3 of 4 models
Structured OutputsAll models

About

GPT-OSS is a family of 4 AI models by OpenAI, released in 2025.

Current Variants

Use-when guidance is based on each model's tracked capabilities, context window, release date, and replacement status.

4 in view
OpenAI GPT OSS Safeguard 120BCurrent

Use when the workload needs safety, 120B parameters, and structured outputs.

2025-09safety120B parametersstructured outputs
gpt-oss-120bCurrent

Use when the workload needs 131k context, 120B parameters, and tool use.

2025-08131k context120B parameterstool use
gpt-oss-20bCurrent

Use when the workload needs 131k context, 20B parameters, and tool use.

2025-08131k context20B parameterstool use
GPT OSS Safeguard 20BCurrent

Use when the workload needs safety, 131k context, and 20B parameters.

2025-08safety131k context20B parameters

Release Timeline

2 release groups
2025-09
1 current
OpenAI GPT OSS Safeguard 120B
safety120B parametersstructured outputs
Current
2025-08
3 current
GPT OSS Safeguard 20B
safety131k context20B parameters
Current
gpt-oss-120b
131k context120B parameterstool use
Current
gpt-oss-20b
131k context20B parameterstool use
Current

Specifications(4 models)

GPT-OSS model specifications comparison
ModelReleasedContextParametersFn CallingTool UseStructured Outputs
OpenAI GPT OSS Safeguard 120B2025-09—120BNoNoYes
gpt-oss-120b2025-08131k120BYesYesYes
gpt-oss-20b2025-08131k20BYesYesYes
GPT OSS Safeguard 20B2025-08131k20BYesYesYes

Pricing

GPT-OSS model pricing by provider
ModelProviderInput / 1MOutput / 1MType
gpt-oss-20bOpenRouter$0.03$0.14Serverless
gpt-oss-120bOpenRouter$0.039$0.18Serverless
gpt-oss-20bNovita AI$0.04$0.15Serverless
gpt-oss-20bVercel AI Gateway$0.05$0.2Serverless
gpt-oss-120bNovita AI$0.05$0.25Serverless
gpt-oss-20bFireworks AI$0.07$0.3Serverless
gpt-oss-20bGCP Vertex AI$0.07$0.25Serverless
GPT OSS Safeguard 20BAWS Bedrock$0.07$0.2Serverless
gpt-oss-20bGroqCloud$0.075$0.3Serverless
GPT OSS Safeguard 20BGroqCloud$0.075$0.3Serverless
GPT OSS Safeguard 20BOpenRouter$0.075$0.3Serverless
GPT OSS Safeguard 20BVercel AI Gateway$0.075$0.3Serverless
gpt-oss-120bGCP Vertex AI$0.09$0.36Serverless
gpt-oss-20bReplicate API$0.09$0.36Serverless
gpt-oss-120bTogether AI$0.15$0.6Serverless
gpt-oss-120bFireworks AI$0.15$0.6Serverless
gpt-oss-120bGroqCloud$0.15$0.6Serverless
OpenAI GPT OSS Safeguard 120BAWS Bedrock$0.15$0.6Serverless
gpt-oss-120bReplicate API$0.18$0.72Serverless
gpt-oss-20bCloudflare Workers AI$0.2$0.3Serverless
gpt-oss-120bCloudflare Workers AI$0.35$0.75Serverless
gpt-oss-120bVercel AI Gateway$0.35$0.75Serverless

Popular comparisons in this family

Frequently Asked Questions

What is GPT-OSS used for?
GPT-OSS is used for safety, agent workflows and tool use, and structured outputs. The family description and listed model capabilities point to those workloads as the best fit.
How does GPT-OSS compare to GPT Realtime 2?
GPT-OSS by OpenAI is strongest where you need safety, while GPT Realtime 2 by OpenAI is the closest related family to check for realtime voice. GPT-OSS has 4 listed variants and reaches up to 131k context, while GPT Realtime 2 reaches up to 131k context, so compare the specs and pricing tables before choosing a production model.
Which GPT-OSS model should I use?
For the lowest listed input price, start with gpt-oss-20b through OpenRouter at $0.03/1M input tokens. For the most capable/latest local choice, evaluate gpt-oss-120b with 131k context and tool use, function calling, and structured outputs.