OpenAI Releases GPT-5.5 and GPT-5.5 Pro to API
OpenAI launched GPT-5.5 on April 23, 2026. This new frontier model handles complex professional work. It became available in the Chat Completions and Responses APIs. The company also rolled out GPT-5.5 Pro for Responses API requests, including Batch. This version tackles tougher problems with additional compute.
GPT-5.5 includes a 1M token context window. It handles image input, structured outputs, function calling, prompt caching, Batch, tool search, built-in computer use, hosted shell, apply patch, Skills, MCP, and web search. The model uses medium reasoning effort by default.
On April 15, the Agents SDK gained new features. These cover running agents in controlled sandboxes, inspecting and customizing the open-source harness, and managing memory creation and storage.
March 2026 Model and API Updates
March 17 brought GPT-5.4 mini and GPT-5.4 nano to Chat Completions and Responses APIs. GPT-5.4 mini delivers GPT-5.4 capabilities in a quicker, more efficient form for high-volume tasks. GPT-5.4 nano focuses on simple high-volume jobs prioritizing speed and cost.
GPT-5.4 mini offers tool search, built-in computer use, and compaction. GPT-5.4 nano provides compaction only, without tool search or computer use.
On March 16, the gpt-5.3-chat-latest slug updated to the current ChatGPT model.
A fix on March 13 addressed an image encoder issue in GPT-5.4 for input_image in Responses and Chat Completions APIs. Image understanding improved for some cases. No user action needed.
March 12 expanded the Sora API. Additions include reusable character references, up to 20-second generations, 1080p output for sora-2-pro, video extensions, and Batch for POST /v1/videos. Sora-2-pro 1080p costs $0.70 per second.
Also on March 12, POST /v1/videos/edits launched for video edits, replacing /v1/videos/{video_id}/remix, which deprecates in 6 months.
On March 5, GPT-5.4 and GPT-5.4 Pro arrived in Chat Completions and Responses APIs. GPT-5.4 suits professional work, with Pro for compute-heavy tasks.
New features included tool search in Responses API to defer tools at runtime, reducing tokens and improving speed. Built-in computer use via Responses API computer tool for UI interaction. A 1M token window and native compaction support longer agents.
March 3 added gpt-5.3-chat-latest to Chat Completions and Responses, matching ChatGPT's GPT-5.3 Instant.
2026 Expansions in February and January
February 24 extended input_file support in Responses and Chat Completions for more document, presentation, spreadsheet, code, and text types.
That day, Responses API gained a phase label for assistant messages as commentary or final_answer.
Also, gpt-5.3-codex joined Responses API.
February 23 launched WebSocket mode for Responses API and gpt-realtime-1.5 for Realtime API, plus gpt-audio-1.5 for Chat Completions.
February 10 added Batch support for GPT Image models: gpt-image-1.5, chatgpt-image-latest, gpt-image-1, gpt-image-1-mini.
Gpt-5.2-chat-latest slug updated to ChatGPT's latest.
Responses API introduced server-side compaction, Skills support for local and hosted execution, and Hosted Shell tool with container networking.
February 9 enabled application/json for /v1/images/edits on GPT image models, using image_url or file_id.
February 3 sped up GPT-5.2 and GPT-5.2-Codex inference by ~40% for API users, without model changes.
January 15 announced Open Responses, an open-source spec for multi-provider LLM interfaces based on OpenAI's Responses API.
January 14 released gpt-5.2-codex to Responses for agentic coding.
January 13 added SIP IP ranges for Realtime API, with GeoIP routing.
Slugs updated: gpt-realtime-mini and gpt-audio-mini to 2025-12-15 snapshots; sora-2 to sora-2-2025-12-08; gpt-4o-mini-tts and gpt-4o-mini-transcribe to 2025-12-15, recommending gpt-4o-mini-transcribe.
January 9 fixed fidelity issue in gpt-image-1.5 and chatgpt-image-latest for image edits.
Key 2025 Releases and Features
December 19 added gpt-image-1.5 and chatgpt-image-latest to Responses API image tool.
December 16 launched those models for advanced image generation.
December 15 released dated audio snapshots: gpt-realtime-mini-2025-12-15, gpt-audio-mini-2025-12-15, gpt-4o-mini-transcribe-2025-12-15, gpt-4o-mini-tts-2025-12-15, with custom voices.
December 11 brought GPT-5.2 to Responses and Chat Completions, improving intelligence, instruction, accuracy, multimodality, code, tools, spreadsheets. New xhigh reasoning, summaries, compaction.
Stay ahead of the AI curve
The most important updates, news, and content — delivered weekly.
No spam. Unsubscribe anytime.
Also, client-side compaction via /responses/compact.
Gpt-5.1-codex-max for long coding tasks.
November 20 added DTMF in Realtime API.
November 13 released GPT-5.1 family to APIs, strong in steerability, code, agents, with none reasoning default. Also gpt-5.1-codex, gpt-5.1-codex-mini; RBAC; extended prompt cache to 24 hours.
October 29 launched gpt-oss-safeguard-120b and -20b.
October 24 added Enterprise Key Management and UK data residency.
October 6 at DevDay: gpt-5-pro, gpt-realtime-mini, gpt-audio-mini, gpt-image-1-mini, v1/videos for Sora 2/Pro, Agent Builder, ChatKit, Trace tools, evals, health dashboard.
October 1 launched IP allowlist.
September 26 added image/file tool outputs in Responses.
September 23 launched gpt-5-codex for Codex CLI.
August 28 made Realtime API generally available.
August 21 added connectors to Responses.
August 20 released Conversations API, migrating from Assistants.
August 7 added GPT-5 family, minimal reasoning, custom tool calls.
Mid-2025 to 2023 API Evolution
June 27 launched Priority processing.
June 24 released o3-deep-research, o4-mini-deep-research; async webhooks, web search.
June 13 added reusable prompts in dashboard/API.
June 10 released o3-pro, reduced o3 prices.
June 4 added fine-tuning with DPO for gpt-4.1 models.
June 3 new snapshots for gpt-4o-audio/realtime previews, Agents SDK.
May 20 added remote MCP, code interpreter to Responses tools; strict mode, schema features.
May 15 launched codex-mini-latest.
May 7 added reinforcement fine-tuning, gpt-4.1-nano fine-tuning.
April 30 launched Enhanced API Budget Alerts.
April 23 added gpt-image-1, updated image endpoints.
April 16 added o3, o4-mini; launched Codex CLI.
April 14 added gpt-4.1 family, 1M context, fine-tuning; deprecated gpt-4.5-preview.
March 20 added audio models to Audio API.
March 19 released o1-pro.
March 11 launched Responses API, tools (web/file search, computer use), Agents SDK, new models; Assistants migration planned.
Earlier months detailed fine-tuning metadata, GPT-4.5 preview, dashboards, data residency, o3-mini, o1 access, admin tools, Usage API, model updates, Realtime features, DevDay announcements, moderation, o1 series, Assistants updates, fine-tuning GA, structured outputs, admin APIs, SSO, GPT-4o mini, uploads, parallel calls, projects, Batch API, GPT-4 Turbo Vision, fine-tuning seeds/checkpoints, Assistants features, and more back to 2023 embeddings, function calling, SDKs.