AI Radar

o1

Deprecated · Language, Multimodal, Reasoning

o1 is an OpenAI reasoning model, presented on 12 September 2024 and released in ChatGPT on 5 December 2024, when ChatGPT Pro launched with o1 pro mode, a version that uses more compute. A post-trained API snapshot, o1-2024-12-17, followed on 17 December 2024 with image input, function calling, Structured Outputs and a reasoning_effort parameter. [1] [2] [3] Primary source

Timeline of o1 →

Claims and evidence

  • Announced Presented with evaluations on 2024-09-12, when only the early version o1-preview was released. Primary source[1]first paragraph [4]first paragraphs
  • Released Released in ChatGPT on 2024-12-05, the day ChatGPT Pro launched with access to o1; the o1 System Card lists an o1-dec5-release checkpoint and the API post of 2024-12-17 refers to the ChatGPT release two weeks earlier. The API snapshot o1-2024-12-17 followed on 2024-12-17 as a new post-trained version. Primary source[2]second paragraph [5]PDF of December 5, 2024, section 2 (o1-dec5-release) [3]OpenAI o1 in the API (model released in ChatGPT two weeks ago)
  • Deprecated Primary source[6]2026-04-22: Legacy GPT model snapshots
  • Released in the API as o1-2024-12-17 (usage tier 5) Primary source[3]OpenAI o1 in the API [7]Dec 17, 2024
  • Status Deprecatedo1 (snapshot o1-2024-12-17) is scheduled to shut down in the API on October 23, 2026. Primary source[6]2026-04-22: Legacy GPT model snapshots
  • Replaced by GPT-5.6 SolSubstitute model listed for o1-2024-12-17 on the deprecations page when checked on 2026-10-01. Primary source[6]2026-04-22: Legacy GPT model snapshots
  • Successor of o1-previewOpenAI calls o1 the successor to o1-preview and the next model in the series. Primary source[3]OpenAI o1 in the API [5]Model data and training
  • Change · Modality Accepts image input in the API, where o1-preview took text only.Compared with o1-preview Primary source[3] [8]
  • Change · Tool use Supports function calling, Structured Outputs and developer messages in the API.Compared with o1-preview Primary source[3]
  • Change · Efficiency OpenAI says o1 uses on average 60 percent fewer reasoning tokens than o1-preview for a given request.Compared with o1-preview Primary source[3]
  • Change · Context length Context window of 200,000 tokens with up to 100,000 output tokens, up from 128,000 tokens for o1-preview.Compared with o1-preview Primary source[9] [8]
  • Input text, image Primary source[9]Model details [3]OpenAI o1 in the API
  • Output text Primary source[9]Model details
  • Feature Function calling Primary source[3]OpenAI o1 in the API [9]Supported features
  • Context window 200,000 tokens Primary source[9]Model details
  • Parameters Not disclosedNot disclosed by OpenAI. No claim madeNo source
  • Access consumer app, API Primary source[2]second paragraph [3]OpenAI o1 in the API

Lineage

Predecessors

Successors

  • o3 · 16 April 2025

Based on

Not derived from another model.

Variants and derived

None recorded.

Siblings

None recorded.

All ancestors

All descendants

Variants

o1 pro mode

  • Released In ChatGPT Pro; released in the API as o1-pro on 2025-03-19. Primary source[2]second paragraph
  • Deprecated Primary source[6]2026-04-22: Legacy GPT model snapshots
  • Variant Inline variant in this record.Version of o1 that uses more compute to think longer, first offered as o1 pro mode in ChatGPT Pro and from 2025-03-19 as o1-pro in the API (Responses API only). Primary source[2]second paragraph [7]Mar 19, 2025 [10]description

Status: Deprecated Evidence not assessed [6]2026-04-22: Legacy GPT model snapshots

Also known as: o1-pro, o1-pro-2025-03-19

Related AI Radar coverage

AI Radar coverage starts in June 2026; no coverage linked yet.

All model releases from OpenAI on AI Radar →