o3 Models by OpenAI

5 models2025Up to 200k ctxFrom $1/1M input

Last refreshed 2026-06-08. Next refresh: weekly.

ResearcherOpenAI

Models5

Released2025

Max context200k

Vision4 of 5 models

Multimodal4 of 5 models

ReasoningAll models

JSON / Tool useAll models

Structured OutputsAll models

Code Execution4 of 5 models

About

o3 is a family of 5 AI models by OpenAI, released in 2025.

Best fit: vision and multimodal workreasoningJSON / Tool use
Capability starting point: o3-pro with 200k context and reasoning, JSON / Tool use, structured outputs, and multimodal inputs
Lowest tracked input: o4-mini · $1/1M · Replicate API
Closest related family: GPT Realtime 2

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

3 in view2 retired

Use when the workload needs 200k context, reasoning, and JSON / Tool use.

2025-10200k contextreasoningJSON / Tool use

o3-proCurrent

Use when the workload needs 200k context, reasoning, and JSON / Tool use.

2025-06200k contextreasoningJSON / Tool use

o3Current

Use when the workload needs 200k context, reasoning, and JSON / Tool use.

2025-04200k contextreasoningJSON / Tool use

Current o3 variants with use-when guidance and lifecycle status
Model	Use when	Released	Signals	Status
o3 Deep Research	Use when the workload needs 200k context, reasoning, and JSON / Tool use.	2025-10	200k contextreasoningJSON / Tool use	Current
o3-pro	Use when the workload needs 200k context, reasoning, and JSON / Tool use.	2025-06	200k contextreasoningJSON / Tool use	Current
o3	Use when the workload needs 200k context, reasoning, and JSON / Tool use.	2025-04	200k contextreasoningJSON / Tool use	Current

4 release groups

2025-10

1 current