GPT-5.5: Gone From ChatGPT, Still in the API at $5 and $30

In brief

GPT-5.5 is OpenAI's flagship model from 23 April 2026, and ChatGPT no longer offers it. OpenAI retired it there on 14 October 2026 but kept it in the API at $5 per million input tokens and $30 per million output, with a 1,050,000-token context window. GPT-6 Sol charges $2 and $10.

Is GPT-5.5 still available?

GPT-5.5 is still available in the API, and nowhere else from OpenAI. The model arrived on 23 April 2026, and OpenAI pulled it from three products on 14 October. The table shows where it stands on 15 November 2026, from the model release notes and the API model page.

A white ceramic OpenAI knot sculpture resting on a low stone plinth in a quiet gallery, with a taller empty plinth behind it and soft window light across the floor.
WhereStatus in November 2026Detail
ChatGPTRetired 14 October 2026Launched for Plus, Pro, Business and Enterprise; GPT-5.5 Pro for Pro, Business and Enterprise
ChatGPT WorkRetired 14 October 2026Same retirement notice
CodexRetired 14 October 2026Launched with a 400K context window for Plus, Pro, Business, Enterprise, Edu and Go
APILiveModel ID gpt-5.5, snapshot gpt-5.5-2026-04-23, 1,050,000 context
API deprecation listNot listedOnly gpt-5.5-cyber-preview has a shutdown date, 1 February 2027
GPT-5.5 lasted 174 days in ChatGPT
GPTGPT-5.4 ThinkingGPT-5.5GPT-5.5 in APIGPT-5.6 SolGPT-6 SolLeft ChatGPT

The 174 days run from 23 April to 14 October, and the count is ours. The dates come from OpenAI's release notes and its launch post, whose update note is dated 24 April for the API. CNBC reported the announcement on Thursday, 23 April.

What did OpenAI say GPT-5.5 could do?

OpenAI called GPT-5.5 "our smartest and most intuitive to use model yet" and aimed it at agentic coding, computer use, knowledge work and early scientific research. It said the model matched GPT-5.4 on per-token speed while using fewer tokens on Codex tasks. Those are OpenAI's own claims, and it made them on the day that the model launched.

The numbers below come from the comparison table in the launch post. The rivals are the models OpenAI picked in April, not today's.

OpenAI's April 2026 tableGPT-5.5GPT-5.4Claude Opus 4.7Gemini 3.1 Pro
Terminal-Bench 2.082.7%75.1%69.4%68.5%
SWE-Bench Pro58.6%57.7%64.3%54.2%
OSWorld-Verified78.7%75.0%78.0%not listed
GDPval (wins or ties)84.9%83.0%80.3%67.3%
BrowseComp84.4%82.7%79.3%85.9%

GPT-5.5 didn't lead everywhere. Claude Opus 4.7 scored higher on SWE-Bench Pro, and Gemini 3.1 Pro scored higher on BrowseComp, in OpenAI's own table. OpenAI also reported 74.0% for GPT-5.5 on the MRCR v2 8-needle test at 512K to 1M tokens, against 36.6% for GPT-5.4. That test checks whether a model can find facts that are buried in a very long prompt, so it matters if you feed it whole codebases.

These scores are from April. GPT-5.6 and GPT-6 have come out since, and OpenAI has not published a refresh of this table for GPT-5.5. So you should treat it as a record of what the model did at the time that it launched, and not as a ranking of what you can use today.

How much does GPT-5.5 cost in the API?

GPT-5.5 costs $5 per million input tokens and $30 per million output tokens in the API. Cached input is $0.50. OpenAI's model page lists a 1,050,000-token context window, 128,000 maximum output tokens and a knowledge cutoff of 1 December 2025.

Per 1M tokens, November 2026InputOutputContext window
GPT-5.5$5.00$30.001,050,000
GPT-5.5 Pro$30.00$180.001,050,000
GPT-6 Sol$2.00$10.001,050,000

GPT-5.5 Pro's row is from its API model page. The GPT-6 Sol row is from its API model page. The context window is the same, and the price is less than half of what it is for GPT-5.5. That is the gap that most of this page comes back to.

There are three rules that change what the bill comes to. Prompts over 272K input tokens are priced at 2x input and 1.5x output for the whole session. Batch and Flex run at half the standard rate, and Priority costs 2.5 times it, per the launch post. Regional data-residency endpoints add 10%.

So a 100,000-token prompt with a 5,000-token answer costs $0.65 on GPT-5.5 and $0.25 on GPT-6 Sol, before caching. That sum is ours, from the list prices above.

How do you still use GPT-5.5 now?

You can still use GPT-5.5 through the API, or through a service that has kept it. OpenAI's retirement notice covers ChatGPT, ChatGPT Work and Codex only. Here's the API route, with the details from the model page.

  1. Create an API key in the OpenAI platform dashboard and add credit.
  2. Send your request to the Responses endpoint, v1/responses, or to Chat Completions, v1/chat/completions.
  3. Set the model to gpt-5.5, or to gpt-5.5-2026-04-23 to lock the snapshot.
  4. Pick a reasoning effort: none, low, medium (the default), high or xhigh.

It needs a developer account and a card, and you pay per token. Plenty of people just want to type a question into a chat box and see what it says. If that's you, Whizi Pro at $29.99 a month lists GPT-5.5 beside Claude and Gemini models, with a first week of $1.99. You get the model in a chat window, not OpenAI's apps. The free plan gives 10 messages a day on Whizi AI, Auto and Claude Haiku 5.5 with no card, and it doesn't include GPT-5.5.

The models still in ChatGPT are covered in ChatGPT model history, and which ChatGPT model should I use says when each one is worth picking.

Should you pick GPT-5.5 or GPT-6 Sol?

Pick GPT-6 Sol for almost everything, because it's the newer model in ChatGPT and costs less per token. OpenAI's 7 October release notes put GPT-6 Sol in ChatGPT on that date, and its API page lists a knowledge cutoff of 20 April 2026, against 1 December 2025 for GPT-5.5.

Keep GPT-5.5 only when there is a reason that you can name. Maybe a pipeline was tuned on its outputs and you need the same behaviour. Maybe you pinned the gpt-5.5-2026-04-23 snapshot for an audit trail. Maybe you're comparing two generations on your own prompts. For a new project, there's no case for paying 2.5 times the input rate.

One caution. OpenAI retired the model from its apps, and it has not announced an API shutdown date for gpt-5.5 itself. That can change at any time, and the deprecations page is where a notice would show up first. So look there before you build anything on a model that OpenAI has already pulled from its own apps, and keep a plan for moving to GPT-6 Sol if a shutdown date ever appears. Don't skip it.

Workflow checklist
  • GPT-5.5 left ChatGPT, ChatGPT Work and Codex on 14 October 2026
  • It is live in the API as gpt-5.5 at $5 in and $30 out per 1M tokens
  • GPT-6 Sol is $2 in and $10 out with the same 1,050,000 context window
  • Check the deprecations page before building anything new on GPT-5.5
  • The April benchmark table is a launch-day record, not a current ranking
Common questions

Frequently asked questions

When was GPT-5.5 released?

OpenAI released GPT-5.5 on 23 April 2026 in ChatGPT and Codex, and added it to the API on 24 April. The API snapshot is named gpt-5.5-2026-04-23. GPT-5.5 Pro launched the same day for Pro, Business and Enterprise users in ChatGPT.

Is GPT-5.5 still in ChatGPT?

No. OpenAI retired GPT-5.5 from ChatGPT, ChatGPT Work and Codex on 14 October 2026. The notice says the change does not affect API access, so the model remains available to developers as gpt-5.5.

What is the GPT-5.5 context window?

The API context window is 1,050,000 tokens, with up to 128,000 output tokens, per OpenAI's model page. Prompts over 272K input tokens cost double on input. In Codex at launch the window was 400K tokens.

How much does GPT-5.5 cost?

In the API, GPT-5.5 costs $5 per million input tokens, $0.50 for cached input and $30 per million output tokens, per OpenAI's model page in November 2026. GPT-5.5 Pro costs $30 and $180. ChatGPT stopped offering GPT-5.5 on 14 October 2026.

GPT-5.5 or GPT-5.6 Sol: which is cheaper?

GPT-5.6 Sol is listed at $4 input and $20 output per million tokens, a promotional price that OpenAI's page says runs until 30 November 2026. GPT-5.5 stays at $5 and $30. The knowledge cutoff is 16 February 2026 for Sol and 1 December 2025 for GPT-5.5.

Is GPT-5.5 better than GPT-6?

No. GPT-6 Sol is the newer model, and OpenAI lists it at $2 input and $10 output per million tokens against $5 and $30 for GPT-5.5. GPT-5.5 scored well in April, but no OpenAI page ranks it above GPT-6.