Documentation 2026-04-22 05:55 UTC to 2026-04-23 05:56 UTC

18 18 

19All deprecations are listed below, with the most recent announcements at the top.19All deprecations are listed below, with the most recent announcements at the top.

20 20 

21### 2026-04-22: Legacy GPT model snapshots

22 

23To improve reliability and make it easier for developers to choose the right models, we are deprecating a set of older OpenAI models. Access to these models will be shut down on the dates below.

24 

25| Shutdown date | Model snapshot | Substitute model |

26| ------------- | ---------------------------------------------------------------------- | --------------------- |

27| 2026-07-23 | `computer-use-preview-2025-03-11` \| `computer-use-preview` | `5.4-mini` |

28| 2026-07-23 | `gpt-4o-audio-preview-2024-12-17` | `gpt-audio` |

29| 2026-07-23 | `gpt-4o-mini-audio-preview-2024-12-17` | `gpt-audio` |

30| 2026-07-23 | `gpt-4o-mini-realtime-preview-2024-12-17` | `gpt-realtime-mini` |

31| 2026-07-23 | `gpt-4o-mini-search-preview-2025-03-11` | `4.1-mini` |

32| 2026-07-23 | `gpt-4o-mini-tts-2025-03-20` | `gpt-realtime` |

33| 2026-07-23 | `gpt-4o-search-preview-2025-03-11` | `gpt-4.1-mini` |

34| 2026-07-23 | `gpt-5-chat-latest` | `gpt-5.3-chat-latest` |

35| 2026-07-23 | `gpt-5-codex` | `gpt-5.4` |

36| 2026-07-23 | `gpt-5.1-chat-latest` | `gpt-5.3-chat-latest` |

37| 2026-07-23 | `gpt-5.1-codex` | `gpt-5` |

38| 2026-07-23 | `gpt-5.1-codex-max` | `gpt-5.4` |

39| 2026-07-23 | `gpt-5.1-codex-mini` | `gpt-5.4-mini` |

40| 2026-07-23 | `gpt-audio-mini-2025-10-06` | `gpt-audio` |

41| 2026-07-23 | `gpt-realtime-mini-2025-10-06` | `gpt-realtime-mini` |

42| 2026-07-23 | `o3-deep-research-2025-06-26` \| `o3-deep-research` | `5.4-Pro` |

43| 2026-07-23 | `o4-mini-deep-research-2025-06-26` \| `o4-mini-deep-research` | `5.4-Pro` |

44| 2026-07-23 | `gpt-5.2-codex` | `gpt-5.4` |

45| 2026-10-23 | `gpt-3.5-turbo-0125` \| `gpt-3.5-turbo`, `gpt-3.5-turbo-completions` | `gpt-4.1-mini` |

46| 2026-10-23 | `gpt-4-0613` \| `gpt-4`, `gpt-4-0613-completions`, `gpt-4-completions` | `gpt-4.1` |

47| 2026-10-23 | `gpt-4-1106-preview` | `gpt-4.1` |

48| 2026-10-23 | `gpt-4-turbo` \| `gpt-4-turbo-2024-04-09`, `gpt-4-turbo-completions` | `gpt-4.1` |

49| 2026-10-23 | `gpt-4.1-nano` \| `gpt-4.1-nano-2025-04-14` | `gpt-5-nano` |

50| 2026-10-23 | `gpt-4o-2024-05-13` | `gpt-4.1` |

51| 2026-10-23 | `gpt-image-1` | `gpt-image-1.5` |

52| 2026-10-23 | `o1-2024-12-17` \| `o1` | `o3` |

53| 2026-10-23 | `o1-pro-2025-03-19` \| `o1-pro` | `5.4-Pro` |

54| 2026-10-23 | `o3-mini-2025-01-31` \| `o3-mini` | `o3` |

55| 2026-10-23 | `ft-o4-mini-2025-04-16` | `gpt-5-mini` |

56| 2026-10-23 | `o4-mini-2025-04-16` \| `o4-mini` | `gpt-5-mini` |

57 

58We are also removing fine-tuned versions of models being deprecated above.

59 

60| Shutdown date | Model snapshot | Substitute model |

61| ------------- | ---------------------------- | ---------------- |

62| 2026-10-23 | `ft-gpt-3.5-turbo` | `gpt-4.1-mini` |

63| 2026-10-23 | `ft-gpt-4` | `gpt-4.1` |

64| 2026-10-23 | `ft-gpt-4.1-nano-2025-04-14` | `gpt-5-nano` |

65| 2026-10-23 | `ft-babbage-002` | `gpt-5-mini` |

66| 2026-10-23 | `ft-davinci-002` | `gpt-5-mini` |

67 

21### 2026-03-24: Sora 2 video generation models and Videos API68### 2026-03-24: Sora 2 video generation models and Videos API

22 69 

23On March 24th, 2026, we notified developers using the Videos API and Sora 2 video generation model aliases and snapshots of their deprecation and removal from the API on September 24, 2026.70On March 24th, 2026, we notified developers using the Videos API and Sora 2 video generation model aliases and snapshots of their deprecation and removal from the API on September 24, 2026.

1---

2latestModelInfo:

3 model: gpt-5.4

4 migrationGuide: /api/docs/guides/upgrading-to-gpt-5p4.md

5 promptingGuide: /api/docs/guides/prompt-guidance.md

6---

7 

1# Using GPT-5.48# Using GPT-5.4

2 9 

3import {10import {

1# Upgrading to GPT-5.4

2 

3# Upgrading to GPT-5.4

4 

5Use this guide when the user explicitly asks to upgrade an existing integration to GPT-5.4. Pair it with current OpenAI docs lookups. The default target string is `gpt-5.4`.

6 

7## Upgrade posture

8 

9Upgrade with the narrowest safe change set:

10 

11- replace the model string first

12- update only the prompts that are directly tied to that model usage

13- prefer prompt-only upgrades when possible

14- if the upgrade would require API-surface changes, parameter rewrites, tool rewiring, or broader code edits, mark it as blocked instead of stretching the scope

15 

16## Upgrade workflow

17 

181. Inventory current model usage.

19 - Search for model strings, client calls, and prompt-bearing files.

20 - Include inline prompts, prompt templates, YAML or JSON configs, Markdown docs, and saved prompts when they are clearly tied to a model usage site.

212. Pair each model usage with its prompt surface.

22 - Prefer the closest prompt surface first: inline system or developer text, then adjacent prompt files, then shared templates.

23 - If you cannot confidently tie a prompt to the model usage, say so instead of guessing.

243. Classify the source model family.

25 - Common buckets: `gpt-4o` or `gpt-4.1`, `o1` or `o3` or `o4-mini`, early `gpt-5`, later `gpt-5.x`, or mixed and unclear.

264. Decide the upgrade class.

27 - `model string only`

28 - `model string + light prompt rewrite`

29 - `blocked without code changes`

305. Run the no-code compatibility gate.

31 - Check whether the current integration can accept `gpt-5.4` without API-surface changes or implementation changes.

32 - For long-running Responses or tool-heavy agents, check whether `phase` is already preserved or round-tripped when the host replays assistant items or uses preambles.

33 - If compatibility depends on code changes, return `blocked`.

34 - If compatibility is unclear, return `unknown` rather than improvising.

356. Recommend the upgrade.

36 - Default replacement string: `gpt-5.4`

37 - Keep the intervention small and behavior-preserving.

387. Deliver a structured recommendation.

39 - `Current model usage`

40 - `Recommended model-string updates`

41 - `Starting reasoning recommendation`

42 - `Prompt updates`

43 - `Phase assessment` when the flow is long-running, replayed, or tool-heavy

44 - `No-code compatibility check`

45 - `Validation plan`

46 - `Launch-day refresh items`

47 

48Output rule:

49 

50- Always emit a starting `reasoning_effort_recommendation` for each usage site.

51- If the repo exposes the current reasoning setting, preserve it first unless the source guide says otherwise.

52- If the repo does not expose the current setting, use the source-family starting mapping instead of returning `null`.

53 

54## Upgrade outcomes

55 

56### `model string only`

57 

58Choose this when:

59 

60- the existing prompts are already short, explicit, and task-bounded

61- the workflow is not strongly research-heavy, tool-heavy, multi-agent, batch or completeness-sensitive, or long-horizon

62- there are no obvious compatibility blockers

63 

64Default action:

65 

66- replace the model string with `gpt-5.4`

67- keep prompts unchanged

68- validate behavior with existing evals or spot checks

69 

70### `model string + light prompt rewrite`

71 

72Choose this when:

73 

74- the old prompt was compensating for weaker instruction following

75- the workflow needs more persistence than the default tool-use behavior will likely provide

76- the task needs stronger completeness, citation discipline, or verification

77- the upgraded model becomes too verbose or under-complete unless instructed otherwise

78- the workflow is research-heavy and needs stronger handling of sparse or empty retrieval results

79- the workflow is coding-oriented, tool-heavy, or multi-agent, but the existing API surface and tool definitions can remain unchanged

80 

81Default action:

82 

83- replace the model string with `gpt-5.4`

84- add one or two targeted prompt blocks

85- read [Prompt guidance for GPT-5.4](https://developers.openai.com/api/docs/guides/prompt-guidance) to choose the smallest prompt changes that recover the old behavior

86- avoid broad prompt cleanup unrelated to the upgrade

87- for research workflows, default to `research_mode` + `citation_rules` + `empty_result_handling`; add `tool_persistence_rules` when the host already uses retrieval tools

88- for dependency-aware or tool-heavy workflows, default to `tool_persistence_rules` + `dependency_checks` + `verification_loop`; add `parallel_tool_calling` only when retrieval steps are truly independent

89- for coding or terminal workflows, default to `terminal_tool_hygiene` + `verification_loop`

90- for multi-agent support or triage workflows, default to at least one of `tool_persistence_rules`, `completeness_contract`, or `verification_loop`

91- for long-running Responses agents with preambles or multiple assistant messages, explicitly review whether `phase` is already handled; if adding or preserving `phase` would require code edits, mark the path as `blocked`

92- do not classify a coding or tool-using Responses workflow as `blocked` just because the visible snippet is minimal; prefer `model string + light prompt rewrite` unless the repo clearly shows that a safe GPT-5.4 path would require host-side code changes

93 

94### `blocked`

95 

96Choose this when:

97 

98- the upgrade appears to require API-surface changes

99- the upgrade appears to require parameter rewrites or reasoning-setting changes that are not exposed outside implementation code

100- the upgrade would require changing tool definitions, tool handler wiring, or schema contracts

101- you cannot confidently identify the prompt surface tied to the model usage

102 

103Default action:

104 

105- do not improvise a broader upgrade

106- report the blocker and explain that the fix is out of scope for this guide

107 

108## No-code compatibility checklist

109 

110Before recommending a no-code upgrade, check:

111 

1121. Can the current host accept the `gpt-5.4` model string without changing client code or API surface?

1132. Are the related prompts identifiable and editable?

1143. Does the host depend on behavior that likely needs API-surface changes, parameter rewrites, or tool rewiring?

1154. Would the likely fix be prompt-only, or would it need implementation changes?

1165. Is the prompt surface close enough to the model usage that you can make a targeted change instead of a broad cleanup?

1176. For long-running Responses or tool-heavy agents, is `phase` already preserved if the host relies on preambles, replayed assistant items, or multiple assistant messages?

118 

119If item 1 is no, items 3 through 4 point to implementation work, or item 6 is no and the fix needs code changes, return `blocked`.

120 

121If item 2 is no, return `unknown` unless the user can point to the prompt location.

122 

123Important:

124 

125- Existing use of tools, agents, or multiple usage sites is not by itself a blocker.

126- If the current host can keep the same API surface and the same tool definitions, prefer `model string + light prompt rewrite` over `blocked`.

127- Reserve `blocked` for cases that truly require implementation changes, not cases that only need stronger prompt steering.

128 

129## Scope boundaries

130 

131This guide may:

132 

133- update or recommend updated model strings

134- update or recommend updated prompts

135- inspect code and prompt files to understand where those changes belong

136- inspect whether existing Responses flows already preserve `phase`

137- flag compatibility blockers

138 

139This guide may not:

140 

141- move Chat Completions code to Responses

142- move Responses code to another API surface

143- rewrite parameter shapes

144- change tool definitions or tool-call handling

145- change structured-output wiring

146- add or retrofit `phase` handling in implementation code

147- edit business logic, orchestration logic, or SDK usage beyond a literal model-string replacement

148 

149If a safe GPT-5.4 upgrade requires any of those changes, mark the path as blocked and out of scope.

150 

151## Validation plan

152 

153- Validate each upgraded usage site with existing evals or realistic spot checks.

154- Check whether the upgraded model still matches expected latency, output shape, and quality.

155- If prompt edits were added, confirm each block is doing real work instead of adding noise.

156- If the workflow has downstream impact, add a lightweight verification pass before finalization.

157 

158## Launch-day refresh items

159 

160When final GPT-5.4 guidance changes:

161 

1621. Replace release-candidate assumptions with final GPT-5.4 guidance where appropriate.

1632. Re-check whether the default target string should stay `gpt-5.4` for all source families.

1643. Re-check any prompt-block recommendations whose semantics may have changed.

1654. Re-check research, citation, and compatibility guidance against the final model behavior.

1665. Re-run the same upgrade scenarios and confirm the blocked-versus-viable boundaries still hold.