'%3e%3crect%20x='16'%20y='51'%20width='68'%20height='27'%20rx='13.5'/%3e%3ccircle%20cx='32'%20cy='53'%20r='15'/%3e%3ccircle%20cx='53'%20cy='43'%20r='20'/%3e%3ccircle%20cx='71'%20cy='57'%20r='12'/%3e%3c/g%3e%3c/svg%3e){kind=small}
Cloud Changelog
OpenAI API changelog
143 announcements since January 2025 · 19 GA · 4 previews · 7 price changes · 6 deprecations
Free. One email a day, only on days when something you follow has an announcement, or one a week if you switch it in your preferences. You confirm your email with a one-time code; there is no password.
October 202615
Eligible Pro and Business subscribers can claim up to $500 monthly in free API creditsPrice
Fast, efficient model for visual reasoning, precise computer use and natural writingGA
Ships with gpt-6-luna, input_json support, and 400 questions per request, up from 200GA
Keeps cached context longer; cache writes cost 2.5x the input token pricePrice
Hosted sandboxes add Artifact SDK and document skills; Ultrafast mode reduces time between tokens
Native Decisions integration lets apps act while the user is speaking
Sandboxes can be pre-warmed before attaching to a session; compute billed only during active turnsPrice
Responses API adds service_tier "ultrafast" to reduce time between output tokens
Snapshot updated to the latest model available in ChatGPT, and will be updated regularly
Beta turns text and images into typed answers with gpt-6-luna, 10x faster than Responses APIPreview
September 202616
Computer use lets agents complete tasks in an OpenAI-hosted browser
Reduces time between output tokens in the Responses API, with US data residency
Fixed an image encoding bug that degraded image understanding, including computer use
Reasoning models with text and image input; Luna costs $0.10 per 1M input tokens
Organization and project admins can restrict or disable new API key creation
Managed Codex harness with durable sessions, MCP server support and OpenAI-hosted or own sandboxesPreview
Expiration dates can be set, with organization or project maximum key lifetime limits
Available in the Responses API for GPT-5.6 and later models, showing reasons for cache missesGA
Offered through trusted access for approved life sciences research; billing starts October 5, 2026GA
Two image models added with xhigh and max quality settings, at GPT Image 2 token rates
Most capable model for reasoning, coding and computer use; drops the none effort levelGA
429 slow_down signals fast traffic growth; 503 server_is_overloaded signals temporary model overload
August 202611
Mutual TLS and X.509 workload identity federation configurable in the Platform consoleGA
whisper-1 and gpt-4o transcribe models shut down February 26, 2027; Assistants API shut downDeprecation
Costs $4 per million input and $20 per million output tokens, promotional through November 21, 2026Price
Regional processing can be selected per request using a prefixed domain
Transparent backgrounds with png or webp output in Images and Responses APIsPreview
New API service tier running up to 14x faster than Standard, for select customersPreview
Two access tiers for approved defenders: Blue with general models, Red with purpose-trained models
Snapshot now points to the latest ChatGPT model for Plus and Pro users
Filtering and grouping by API key, also in Usage and Costs APIs
July 20266
June 20268
Snapshot now points to the latest Instant model used in ChatGPT
Shows blocked Responses requests by safety_identifier values sent to identify end users
Billed per minute with 5-minute minimum from June 2, 2026, instead of 20-minute ratePrice
Available through an OpenAI-compatible Responses API endpoint; support varies by Region
May 202613
prompt_cache_retention defaults to 24h instead of in_memory for organizations without ZDR
Points to the latest Instant model used in ChatGPT, regularly updated
Workloads exchange external identity tokens for short-lived access tokens, no long-lived API keysGA
Multiple IP allowlists can be applied at project or organization level
dall-e-2, dall-e-3 and Realtime API Beta removed on May 12, 2026Deprecation
Adds return_token_budget to opt in to longer GPT-5+ reasoning web search runs
Plugin gives Codex OpenAI Platform access and API setup guidance for building apps and agents
April 20263
March 20267
Released in Chat Completions and Responses API; mini supports tool search and computer useGA
Slug now points to the latest model currently used in ChatGPT
Fixes a bug with input_image inputs; some image understanding quality may improve
POST /v1/videos/edits added; /v1/videos/{video_id}/remix to be deprecated in 6 monthsDeprecation
New frontier models with tool search, computer use, 1M-token context and CompactionGA
February 202612
Accepts more document, presentation, spreadsheet, code and text file types
GPT-Realtime-1.5 added to Realtime API, gpt-audio-1.5 to Chat CompletionsGA
Responses API adds Skills, with local and hosted container execution
Hosted Shell tool launched, with networking support in containers
gpt-5.2-chat-latest slug now points to the model currently used in ChatGPT
Batch API now supports GPT Image models, including gpt-image-1.5
/v1/images/edits accepts application/json requests for GPT image models
GPT-5.2 and GPT-5.2-Codex now run about 40% faster, weights unchanged
January 20267
Open Responses spec released, an open-source standard for interoperable LLM interfaces
gpt-5.2-codex released in the Responses API for agentic coding
Realtime API adds dedicated SIP IP ranges with GeoIP routing
gpt-4o-mini-tts and gpt-4o-mini-transcribe slugs now point to 2025-12-15 snapshots
gpt-realtime-mini and gpt-audio-mini slugs now point to 2025-12-15 snapshots
Fixed gpt-image-1.5 using high fidelity for image edits when set to low