AI Assistant Roadmaps: Reading Model Release and Retirement Notes

Vendor release and retirement notes identify model status and scheduled shutdowns, but they do not show which assistant fits your workflow.

Sources checked 2 Oct 2026

The vendors’ own release and deprecation pages can establish a documented product surface, status, and scheduled retirement date; they cannot establish whether an assistant fits your workflow. Chat Picker has not tested, benchmarked, scored, or ranked ChatGPT, Claude, or Gemini, so this guide reads the official pages and shows how to verify the remaining questions.

What the vendors document

Read status and surface first. Release notes record announced changes; deprecation pages describe lifecycle status. A consumer-app change does not establish API availability, and an API retirement does not automatically describe the consumer app.

As read on October 1, 2026, OpenAI’s API deprecations page gives generally available models at least six months’ notice and specialized generally available variants at least three months. Preview notice may be as short as two weeks; safety or compliance concerns can require earlier retirement. Legacy models and endpoints receive no updates and become inaccessible at shutdown.

OpenAI’s page schedules API removal for gpt-5.4-cyber on October 1, 2026; older GPT Image models on December 1, 2026; legacy audio, realtime, and transcription families on January 20, 2027; and whisper-1, gpt-4o-transcribe, gpt-4o-mini-transcribe, and gpt-4o-transcribe-diarize on February 26, 2027. It also deprecates reusable prompt objects and Agent Builder, while ChatKit remains available.

OpenAI’s ChatGPT release notes, read on the same date, record rolling-out Library connections to Box, Dropbox, and SharePoint alongside Google Drive, plus a Privacy Center rollout. Those entries do not state calendar launch dates.

As read on October 1, 2026, Anthropic’s model deprecations page distinguishes active, legacy, deprecated, and retired models. A deprecated model remains functional but is no longer recommended and has a replacement and retirement date; requests to a retired model fail. Claude Opus 4.1 was retired on August 5, 2026. Developers using Claude Sonnet 4.5 were notified on September 30, 2026, of an upcoming API retirement, but the page does not state a shutdown date.

Anthropic says active deployments of publicly released models receive at least 60 days’ notice. Its listed dates cover the Claude API, Claude Platform on AWS, and Microsoft Foundry; Amazon Bedrock and Google Cloud set their own schedules.

Claude’s release notes, read on October 1, 2026, record that Claude Opus 5.5 costs 40% less to run than Opus 5. They also place Claude Design, Claude Slides, and Claude Docs artifacts on every plan, including Free, although Enterprise versions are beta and off by default. These entries do not state launch dates.

As read on October 1, 2026, Gemini’s API deprecations page says the 2.5 models are not deprecated and will remain available until further notice, although access is limited to users who have actively used them. Google recommends 3.5 Flash-Lite or 3.8 Flash for new projects. Its listed shutdown dates are the earliest possible dates; Google says it will communicate the exact date separately.

Gemini’s API changelog, read on October 1, 2026, records introductory pricing for Gemini 3.7 Flash through December 31, 2026, Gemini 3.8 Live as the default for most low-latency voice-agent experiences, and the Gemini Deep Research agent as a preview launch. These entries do not state calendar launch dates. Gemini Apps’ release updates separately record Gemini 3.1 Pro rolling out globally and Temporary Chat conversations not being saved to chat history.

Roadmap status does not settle plan fit. As read on October 1, 2026, OpenAI’s Plus help page lists Plus at $20 a month; Anthropic’s pricing page lists Pro at $20 monthly or $17 a month with annual billing and $200 upfront; and Google’s US plans page lists Google AI Pro at $19.99 a month.

The ChatGPT, Claude, and Gemini free-plan pages differ: OpenAI limits uploads, images, voice, and deep research; Anthropic lists web search, file creation, code execution, and memory; Google requires an Account and lists model access and storage. The ChatGPT and Claude pricing pages give no exact message counts. Google describes compute-based limits that refresh every five hours up to a weekly limit.

For data controls, OpenAI lists an opt-out from training on chats for Free, Go, Plus, and Pro. Anthropic lists an opt-out for Free, Pro, and Max and says Team is not trained on by default. Google’s plans page links to data-handling information but states no training setting.

On accuracy, OpenAI says outputs can be incorrect or misleading and important information should be checked. Anthropic says not to rely on Claude as the only source of truth and to inspect original sources. Google’s source page explains how to open the Sources panel but states no general reliability limit.

What the documentation cannot tell you

These pages cannot show how an assistant will handle your instructions, documents, tools, or source synthesis. They also cannot establish migration compatibility, whether a model is exposed through your chosen plan or platform, or the cost of your actual workflow. Treat a roadmap entry as documented status, then verify the product surface your account can access.

How to check it yourself

Run the same checks in each assistant, using a fresh conversation where appropriate.

Read one lifecycle entry. Give the assistant a copied vendor entry. Ask: “State the exact product or model name, status, replacement, notice period, and shutdown date. Quote the source and say ‘not stated’ for absent fields.” Look for scheduled versus completed retirement. Record every omission or added claim.
Separate app and API. Give it one consumer release entry and one API deprecation entry. Ask: “For each item, name the documented surface, quote what changed, and list only dates printed in the entry.” Look for explicit app or API wording. Record any inference or missing date.
Check plan fit. Paste the current plan table. Ask: “For the plan named above, list its price, included limits, context window, tools, and file limits. Mark every unstated field ‘not stated.’” Look for a current source and a distinction between dollar rates and included allowances. Record unknowns.
Check data controls. Paste the account privacy text. Ask: “List documented controls for training on chats, connected files, and retention. Quote each control and its source; do not infer account settings.” Look for plan-specific wording. Record controls your account does not actually expose.
Check evidence. Ask: “Using OpenAI’s API deprecations page, what is the scheduled API removal entry for gpt-5.4-cyber? Give the exact name, date, and direct link, and distinguish a scheduled removal from a completed shutdown.” Compare the answer with the page and record mismatches.
Check migration needs. Tell the assistant: “A deployed integration calls Claude Sonnet 4.5 through the Claude API and cannot change models on launch day. Using Anthropic’s deprecations page, make a pre-retirement migration checklist. Mark unknown dates or replacement models as not stated.” Record dependencies, test cases, rollback steps, and unsupported additions.

Which rows of the comparison matter

In the Claude vs ChatGPT vs Gemini matrix, check Free plan, Low-cost tier, Main paid plan, Heavy-use plans, and Team plan when relevant. Then compare Context window in the app, How usage limits are described, Training on your chats, and Ads.

Each published figure carries its vendor source and read date. If a figure could not be confirmed, the matrix leaves it blank and marks it “Not verified.” Use the roadmap and deprecation pages afterward to confirm the exact product surface and lifecycle status.