AI Pricing This Week: OpenAI Shutdown, Mistral, Oct 19-25

Short answer. From October 19 to 25, 2026, seven models we track shut down: Gemini Omni Flash Preview and six OpenAI models. Mistral added Mistral Large 4 at $1.36 / $4.18 per million tokens. No provider changed a list price in our data. We corrected four of our own readings, the largest a table on OpenAI’s pricing page we had missed.

In plain words. Some old models stopped working, and a few new ones arrived. We also fixed mistakes in how we had read OpenAI’s and Mistral’s price pages. Prices here are per million tokens. A token is a short piece of text, about three-quarters of a word.

This is the second weekly summary. It covers Monday, October 19 to Sunday, October 25, 2026, plus what we read again on Monday, October 26, which is marked as such. Every fact comes from AI Signal’s dataset and change log, with the provider page it was read on and the date.

The week in three lines. Two shutdowns took effect: Gemini Omni Flash Preview on October 22 and six OpenAI models on October 23, and Together AI took two open models off serverless. Our largest correction was OpenAI’s folded “All models” table, which our reader did not see; 14 OpenAI rows entered the dataset on October 26 because of it. Mistral added two models, one its own and one made by Z.ai.

Shutdowns

Gemini Omni Flash Preview, October 22. Google’s deprecations page lists “gemini-omni-flash-preview | June 30, 2026 | October 22, 2026 | gemini-omni-1.1-flash |”, the last field being the replacement. On October 23 Google’s model page printed “Model versions: Stable: gemini-omni-1.1-flash Shut down: gemini-omni-flash-preview”, and the pricing page no longer listed it. Its page stays, marked as shut down.

Six OpenAI models, October 23. o1, o1-pro, o3-mini, GPT-4, GPT-4 Turbo and GPT-3.5 Turbo. Five point to GPT-5.6 Sol, for example “October 23, 2026 | o1-2024-12-17 | o1 | gpt-5.6-sol |”, and GPT-3.5 Turbo to GPT-5.6 Terra. Read again on October 26 in its Markdown version, OpenAI’s deprecations page has 17 rows dated “October 23, 2026”: 12 in the table under “2026-04-22: Legacy GPT model snapshots”, which says “Access to these models was shut down on October 23, 2026.”, and five fine-tuned versions below it. The 12 include models we do not track, such as gpt-4-1106-preview, gpt-4.1-nano and gpt-image-1. On the same day, 9 of those 12 still had a row on OpenAI’s pricing page, for example “o1-pro | $150.00 | $600.00 | - |”. A shut-down model cannot be called whatever its row says, so our dataset marks the six as retired, and the pricing table no longer shows them. The cost of moving is in OpenAI’s October 23 shutdown.

Two open models on Together AI, October 22. Together’s deprecations page lists “2026-10-22 | meta-llama/Llama-3.3-70B-Instruct-Turbo | meta-models/Muse-Glimmer-30B | Yes |” and “2026-10-22 | deepseek-ai/DeepSeek-V4-Flash-0731 | deepseek-ai/DeepSeek-V4.1-Flash | Yes |”, read on October 26. Each model now has one current host, so Llama 3.3 70B Instruct and DeepSeek V4 Flash 0731 are kept as records, with Together’s offer listed apart from the price table.

Our own corrections

None of these was a price change by a provider. Each has an entry in the price-change log that says so in its first line.

OpenAI’s “All models” table, October 26. OpenAI’s pricing page has a folded “All models” table that the page draws in the browser. The text our reader took from the page did not include it. Its Markdown version, developers.openai.com/api/docs/pricing.md, prints it. Two statements were wrong because of that:

• From September 3, 2026, we showed GPT-5, GPT-5 mini and o3 with no price and said GPT-5’s price was no longer published. The table, read on October 26, prints “gpt-5 | $1.25 | $10.00 | $0.125”, “gpt-5-mini | $0.25 | $2.00 | $0.025” and “o3 | $2.00 | $8.00 | $0.50”. The three rows carry those prices again. All three shut down on December 11.
• On October 5, we wrote that OpenAI’s pricing page no longer listed the six models above; the post about their shutdown repeated it. On October 26 the table still listed them, at the prices we carry. The post now has a dated note, and the log entry of October 5 is followed by one that corrects it.

From the same table we added 14 rows OpenAI already priced (see New models below) and their Batch, Flex and Fast mode figures. Since October 26 the weekly price check and the daily new-model scan read the Markdown version of a pricing page as well as its text, where the provider publishes one, and a priced row missing from the text is added from it.

Mistral’s cached input, October 22. Mistral’s pricing page prints a cached-input price in a “Cached input” column, and seven of our Mistral rows did not carry it: Mistral Large 3 ($0.05), Mistral Medium 3.5 ($0.15), Mistral Small 4 ($0.015), Ministral 3 14B ($0.02), Ministral 3 8B ($0.015), Ministral 3 3B ($0.01) and Codestral ($0.03) per million tokens. Input and output prices did not change. When Mistral began printing these figures is not known.

A quote that lost “$1”, October 26. When the two notes for Mistral Large 3 and Mistral Medium 3.5 were written on October 22, a script dropped “$1” from the quoted rows, so the cell that the page prints as “$1.5” came out wrong. The prices were right; the quotes are now “Mistral Large 3 ↗ | $0.5 | $0.05 | $1.5 |” and “Mistral Medium 3.5 ↗ | $1.5 | $0.15 | $7.5 |”, and a test now fails on that shape in any data file.

Muse Spark’s context window, October 26. We stored 1,000,000 tokens for Muse Spark 1.1, 1.2 and 1.3. Meta’s models documentation says “All take text, image, video, audio, and PDF input, return text, and have a 1,048,576-token context window.” The three rows now carry 1,048,576.

New models

Update, October 30, 2026: Mistral has since cut Mistral Large 4’s price. Its pricing page printed Mistral Large 4 ↗ | $0.7 | $0.07 | $2.1 | on October 30, 2026, against the $1.36 / $4.18 in the table above, the price it showed when AI Signal read it on October 22 and 26, 2026. See Mistral Large 4’s price cut.

Mistral’s model page prints “October 6, 2026 Public Preview Open v 26.10” for Mistral Large 4, and its lifecycle page says Public Preview models “Are priced at the same rate as General Availability models.” GLM 5.3 Flash is made by Z.ai; Mistral’s page says “October 22, 2026 Blog License GA Third-party Open v 5.3-flash” and “The model is served without Mistral modifications.” Each has its own post: Mistral Large 4 API pricing and GLM 5.3 Flash API pricing.

On October 26, from OpenAI’s “All models” table, we added five current rows: GPT-5.4 mini at $0.75 / $4.50, GPT-5.2 at $1.75 / $14.00, GPT-5.2 Pro at $21.00 / $168.00, GPT-5.5 Cyber at $12.50 / $75.00 and GPT-5 Search API at $1.25 / $10.00. They are not new at OpenAI, which released GPT-5.2 on December 11, 2025; they were missing here. Nine deprecated rows entered with them: GPT-4.1, GPT-4.1 mini, GPT-4o, GPT-4o mini, GPT-5 nano, GPT-5 Pro, GPT-5.1, GPT-5.4 nano and o3-pro.

Subscriptions and coding tools

GitHub’s Copilot plans page, read on October 23, no longer says “50 chat requests (including Copilot Edits)”; it now says “GitHub Copilot Free users are limited to 2,000 completions and access to entry level models per month.” Prices did not change. Claude Code’s page, read in a browser on October 26, shows Max 5x and Max 20x on one card: “Claude Code is included in your Max plan. Choose 5x or 20x usage for everyday work in larger codebases.” Its prices did not change either. On October 20 we corrected a line of our own: “Includes Gemini 4 Argon” is the label above Google AI Pro’s card, not a feature of Google AI Plus.

Fireworks’ page for Muse Spark Open, read on October 26, no longer prints the limited-preview sentence that named October 29. It prints “Available on serverless on US-only capacity at standard serverless pricing, with no US-only premium.” at the same $1.25 / $4.25.

Other posts of the week: Microsoft Copilot vs ChatGPT Plus, free AI coding tools, Claude Code vs OpenAI Codex and Grok vs ChatGPT in Spain.

New pages

Three maker pages went up on October 26: Mistral, xAI and Meta. Mistral’s page lists the two Z.ai models it serves apart from its own.

What comes next

Dates as the providers print them, all on Deprecations:

• November 30: Anthropic retires Claude Sonnet 4.5: “November 30, 2026 | claude-sonnet-4-5-20250929 | claude-sonnet-5-5 |”. The same day is the last of Gemini 4 Argon’s introductory price.
• December 11: OpenAI shuts down GPT-5, GPT-5 mini, GPT-5 nano, GPT-5 Pro, o3 and o3-pro. Its page, read on October 26, lists “Dec 11, 2026 | gpt-5 | gpt-5-2025-08-07 | gpt-5.6-sol |” and five rows like it.
• April 1, 2027: OpenAI removes GPT-5.1, GPT-5.4 nano and GPT-5.3 Codex: “The following models are deprecated and will be removed from the API on April 1, 2027, with six months’ notice.”
• GLM 5.2 on Mistral: deprecated, with no retirement date on its model page, which reads “Deprecation date 9/29/2026” on October 26. We will read it again on November 1.

How we checked

The facts were listed by our weekly script from AI Signal’s own data files: the models dataset, the price-change log, the deprecation records, the subscription and coding-tool data and the posts of the week. Every price and date was read on the provider’s own page, on the date given, October 26 included. Model prices are list prices in US dollars per million tokens. We do not know why any of these prices are what they are, and this summary does not guess.

Frequently asked questions

Which AI models shut down between October 19 and 25, 2026?

Gemini Omni Flash Preview on October 22, and OpenAI's o1, o1-pro, o3-mini, GPT-4, GPT-4 Turbo and GPT-3.5 Turbo on October 23, along with six other OpenAI models and five fine-tuned versions that AI Signal does not track. On October 22 Together AI also removed Llama 3.3 70B Instruct Turbo and DeepSeek V4 Flash 0731 from serverless.

What did AI Signal correct about its own prices that week?

Our reader did not see the folded All models table on OpenAI's pricing page, so GPT-5, GPT-5 mini and o3 had shown no price since September 3; they are back at $1.25 / $10.00, $0.25 / $2.00 and $2.00 / $8.00. Seven Mistral rows also gained their cached-input price on October 22 and two Mistral quotes were fixed on October 26, and none of the four was a price change by the provider.

Which new models were added that week?

Mistral Large 4, in Public Preview at $1.36 input and $4.18 output per million tokens, and Z.ai's GLM 5.3 Flash on Mistral at $0.15 and $0.50. On October 26, fourteen OpenAI rows that OpenAI already priced were added: five current models and nine deprecated ones.