Monthly AI Changes

A dated monthly log of what changed, why it matters, and where to verify it. Newest entries first, each linked to the vendor's own page.

Newest entry logged: October 23, 2026. October and late-September entries were checked against vendor pages on that date.

How this changelog works

AI vendors ship changes weekly. Pricing shifts, model names get renamed, capabilities get deprecated, and policy pages get rewritten without announcement. If you build workflows on top of these platforms, you need a single place to track what actually changed: not marketing claims, and not rumor.

This page is a dated review record, not a live feed. We log the changes that matter for people running AI in production: pricing updates, plan restructures, regulatory milestones, and model releases that change what you can build. Every entry links to the vendor's own source page so you can verify the claim yourself.

We do not chase every model release. We log the changes that force a decision: a plan you need to re-evaluate, a deprecation that breaks a workflow, a compliance date that changes your roadmap. If a change shows up here, treat it as worth a 10-minute review at your next planning meeting.

October 2026

  • October 23: OpenAI's deprecations page lists October 23, 2026 as the shutdown date for a batch of legacy API models, including gpt-3.5-turbo, gpt-4, gpt-4-turbo, gpt-4.1-nano, o1, o1-pro, o3-mini, o4-mini and gpt-image-1. Google lists the same date for deep-research-pro-preview-12-2025.
  • October 23 (observed): OpenAI's help center says ChatGPT agent is no longer available and points to ChatGPT Work and the cloud browser. Microsoft's support page says Copilot Pro is no longer sold (support ended August 1, 2026); Microsoft now lists Microsoft 365 Pro at $99.99/month (€110.00 in Germany).
  • October 21: Anthropic released API version 2026-10-21, which removes temperature, top_p and top_k (requests that send them get a 400 error). Claude Duet opened as a US-only research preview for Pro and Max. ChatGPT ads expanded to 89 countries (Free and Go only).
  • October 15: Claude Opus 5.5 fast mode ($8 / $40 per million tokens) left research preview. Meta released Muse Spark Open weights (816B total, 42B active parameters, Muse Model License). Anthropic's tentative "not sooner than October 15, 2026" retirement date for Claude Haiku 4.5 passed with no deprecation notice; it is still listed as active.
  • October 14: OpenAI retired GPT-5.5 from ChatGPT, ChatGPT Work and Codex. API access was not affected: gpt-5.5 is still priced at $5 / $30 with no shutdown date. OpenAI also deprecated the GPT-4o, GPT-4o mini, GPT-4.1 and GPT-4.1 mini snapshots, with shutdown listed for January 14, 2027.
  • October 13: Google made Gemini 4 Argon generally available as its flagship model at $2 / $10 per million tokens through November 30, 2026, then $4 / $20 from December 1, 2026. GitHub added it to Copilot the same day.
  • October 12: xAI launched X Subscription (Plus $8, Premium $30, Super $100, Ultra $200). Premium bundles SuperGrok, Cursor Pro and X Premium+, and Cursor's pricing page now sells through X Subscription. Meta launched Muse Wonder on its Model API at $4 / $20.
  • October 9: Google stopped selling new Gemini Code Assist Standard and Enterprise subscriptions; existing ones renew through 2026 and Google points to Antigravity.
  • October 8: Google deprecated Gemini 3.7 Flash and Gemini 3.5 Flash, auto-routing requests to Gemini 3.8 Flash and Gemini 3.6 Flash.
  • October 7: GPT-6 became the ChatGPT Chat model (GPT-6 Sol on Plus, Pro, Business and Enterprise; GPT-6 Luna on Free and Go from October 8, per OpenAI's October 7 release notes). Anthropic released Claude Haiku 5.5 ($0.10 / $0.50 for prompts up to 100K tokens), cut Sonnet 5.5 cache reads to $0.10, and added monthly API credits to Max ($100 or $200) and Team (up to $500 pooled).
  • October 6: Mistral Large 4 entered public preview with a 1M-token context at $1.36 / $4.18. Google made Nano Banana 2.1 generally available.
  • October 1: OpenAI deprecated gpt-5.3-codex, gpt-5.1 and gpt-5.4-nano, with shutdown listed for April 1, 2027.
https://developers.openai.com/api/docs/deprecationshttps://help.openai.com/en/articles/9624314-model-release-noteshttps://help.openai.com/en/articles/6825453-chatgpt-release-noteshttps://help.openai.com/en/articles/11752874-chatgpt-agenthttps://openai.com/index/chatgpt-ads-expands-26-new-countries/https://platform.claude.com/docs/en/release-notes/overviewhttps://platform.claude.com/docs/en/about-claude/model-deprecationshttps://support.claude.com/en/articles/12138966-release-noteshttps://ai.google.dev/gemini-api/docs/changeloghttps://ai.google.dev/gemini-api/docs/pricinghttps://github.blog/changelog/label/copilot/https://x.ai/news/x-subscriptionhttps://cursor.com/pricinghttps://docs.cloud.google.com/gemini/docs/codeassist/release-noteshttps://support.microsoft.com/en-us/microsoft-365-copilot/about-microsoft-copilot-prohttps://docs.mistral.ai/resources/changelogshttps://research.meta.ai/blog/muse-spark-open

September 2026

  • September 30: Anthropic deprecated Claude Sonnet 4.5 and lists November 30, 2026 as its retirement date, with Claude Sonnet 5.5 as the replacement.
  • September 29: OpenAI released GPT-6.1 Sol ($2 / $10 per million tokens) and added GPT-6 Astra Ultrafast. In ChatGPT it launched Pro $500 with Ultrafast, reopened Pro $200 to new subscribers with a lower allowance, and introduced dots always-on agents for Pro and Business Premium (Pro access excluded the EEA, Switzerland and the UK at launch).
  • September 28: Anthropic released Claude Sonnet 5.5 at $2 / $10 per million tokens.
  • September 22: Anthropic released Claude Opus 5.5 at $4 / $20 per million tokens with a 1M-token context window and up to 128K output tokens. OpenAI released GPT-6 Sol ($2 / $10) and GPT-6 Luna ($0.10 / $0.50). xAI announced Grok 4.7 ($2 / $6 under 200K prompt tokens).
  • September 15: Meta announced Meta One subscriptions (Core $7.99, Premium $19.99), with basic Meta AI staying free.
  • September 3: OpenAI released GPT-6 Astra at $10 / $50 per million tokens with a 1.05M-token context window.
  • September 2: Google made Gemini 3.8 Flash generally available at $0.75 / $3.75 through December 31, 2026, then $1.50 / $7.50 from January 1, 2027.
  • September 1: Anthropic released Claude Fable 5.1 at $10 / $50 per million tokens. Claude Sonnet 5's $2 / $10 price, first announced as introductory, had already become the standard price on August 10; the planned September 1 increase to $3 / $15 did not happen.
  • September (announced): OpenAI said GPT-5.5 would retire from ChatGPT, ChatGPT Work and Codex on October 14, 2026. It did; see October.
https://platform.claude.com/docs/en/about-claude/model-deprecationshttps://developers.openai.com/api/docs/changeloghttps://developers.openai.com/api/docs/modelshttps://help.openai.com/en/articles/6825453-chatgpt-release-noteshttps://help.openai.com/en/articles/9793128-about-chatgpt-pro-tiershttps://platform.claude.com/docs/en/release-notes/overviewhttps://platform.claude.com/docs/en/about-claude/pricinghttps://docs.x.ai/docs/modelshttps://about.fb.com/news/2026/09/introducing-meta-one-subscription-service-more-features-ai/https://ai.google.dev/gemini-api/docs/changelog

August 2026

  • Article 50 transparency obligations under the EU AI Act began applying on August 2, 2026. The standalone high-risk (Annex III) obligations originally scheduled for the same date were moved to December 2, 2027 by the Digital Omnibus on AI.
  • OpenAI reduced GPT-5.6 Sol API pricing to $4 input / $20 output per million tokens on August 21, 2026.
https://eur-lex.europa.eu/eli/reg/2026/1744/ojhttps://developers.openai.com/api/docs/changelog

July 2026

  • Regulation (EU) 2026/1744, the Digital Omnibus on AI, entered into force on July 27, 2026. It moves Annex III high-risk obligations to December 2, 2027 and Annex I obligations to August 2, 2028; Article 50 transparency still applies from August 2, 2026.
  • OpenAI released the GPT-5.6 family (Sol, Terra, and Luna) on July 9, 2026.
  • Anthropic listed Claude Opus 5 as its premium Opus model, with a 1M-token context window and $5 input / $25 output API pricing per million tokens.
  • Anthropic listed Claude Fable 5 as generally available for demanding reasoning and long-horizon agent work at $10 input / $50 output per million tokens.
  • OpenAI's ChatGPT plan page reflected the GPT-5.6-era plan surface at the time. Confirm exact entitlement limits on the linked page before purchasing.
https://eur-lex.europa.eu/eli/reg/2026/1744/ojhttps://developers.openai.com/api/docs/changeloghttps://platform.claude.com/docs/en/about-claude/models/overviewhttps://platform.claude.com/docs/en/about-claude/models/introducing-claude-fable-5-and-claude-mythos-5https://chatgpt.com/pricing/

June 2026

  • GitHub moved Copilot to usage-based billing with AI Credits on June 1, 2026. Plan prices stayed the same: Pro $10, Pro+ $39, Business $19 per user, Enterprise $39 per user.
  • Claude Sonnet 5 became available to API customers on June 30, 2026 with introductory $2 input / $10 output pricing per million tokens through August 31, 2026. That price later became the standard price.
  • Anthropic published the Fable 5 and Mythos 5 model introduction, including Fable 5 availability and pricing details.
https://github.blog/news-insights/company-news/github-copilot-is-moving-to-usage-based-billing/https://platform.claude.com/docs/en/about-claude/models/whats-new-sonnet-5https://platform.claude.com/docs/en/about-claude/models/introducing-claude-fable-5-and-claude-mythos-5

May 2026

  • Anthropic published Claude Opus 4.8 information before its current model overview listed Claude Opus 5 as active.
  • The AI plan comparator retained source links for OpenAI, Anthropic, Google, and Perplexity while its full consumer-plan audit was queued.
https://www.anthropic.com/news/claude-opus-4-8https://openai.com/chatgpt/pricing/https://claude.com/pricing

April 2026

  • OpenAI published GPT-5.5 information for ChatGPT and API users before the GPT-5.6 and GPT-6 releases.
  • The EU AI Act implementation timeline continued toward its August 2026 milestone.
https://openai.com/index/introducing-gpt-5-5/https://digital-strategy.ec.europa.eu/en/policies/regulatory-framework-ai

March 2026

  • OpenAI pricing page reflects latest plan structure and model access tiers.
  • Perplexity plan guidance updated with Standard/Pro/Max and enterprise options.
  • EU AI Act implementation timeline remains active with August 2026 milestone approaching.
https://openai.com/chatgpt/pricing/https://www.perplexity.ai/help-center/en/articles/11187416-which-perplexity-subscription-plan-is-right-for-youhttps://digital-strategy.ec.europa.eu/en/policies/regulatory-framework-ai

February 2026

  • Anthropic released Claude Sonnet 4.6 update information.
  • DeepSeek published updated pricing guidance for V3.2 APIs.
https://www.anthropic.com/news/claude-sonnet-4-6https://api-docs.deepseek.com/quick_start/pricing/

January 2026

  • Google AI plan pages continued clarifying AI Pro and premium capabilities.
  • Microsoft Copilot pricing page continued listing business plan structure and add-ons.
https://one.google.com/about/planshttps://www.microsoft.com/en-us/microsoft-365-copilot/pricing

What qualifies as a change

Not every vendor update makes this log. We include a change only when it meets one of four tests.

  • Pricing or plan structure shifts: a new tier, a renamed plan, a seat minimum change, or a capability moved behind a paywall.
  • Model availability changes: releases, deprecations, rate-limit shifts, or API pricing updates that affect production cost models.
  • Policy or compliance updates: regulatory milestones, terms-of-service changes, data-handling shifts, or vendor-published governance guidance.
  • Capability or integration changes: new MCP connectors, workspace features, or agent capabilities that change what you can ship.

We do not log benchmark wins, marketing launches, or blog posts that don't map to an actual product change. The bar is: if a reader acts on this entry, something in their stack, budget, or plan will be different.

How to use this page

Treat this as a monthly review input. At the start of each month, scan the latest entries against your current AI spend, your active model contracts, and any compliance commitments you've made. Flag anything that requires a decision and move it into your ops review.

Every linked source is the vendor's own page, not a third-party summary. If you're presenting a finding to a stakeholder, quote the vendor source directly; that's what they'll ask for when you propose a change.

FAQ

How often is this page updated?

We aim to add a dated material-change entry each month after a source audit. When a critical mid-month change is verified, we add it as soon as practical and note the date.

Why don't you include every model release?

Because most releases don't force a decision. We include releases that unlock new capability classes, deprecate existing endpoints, or change pricing. A slightly-better benchmark score is not a change worth logging.

How do I verify an entry?

Click the source link under each entry. Every claim on this page resolves to the vendor's own documentation, pricing page, or official announcement. If a vendor page has changed since we logged the entry, the archived version is usually retrievable via the Wayback Machine.