Reference · updated monthly

What's current: models, plans, changes

The only place on the site that holds prices and version numbers. Lessons explain principles; specific numbers are checked here. Every entry has a check date and a link to its official source.

Checked: October 2026

Assistants and models

Claude

Anthropic · claude.ai

Claude in the tool catalog

Models

Claude Opus 5.5 — The main high-end model: long work with code and documents. The default model in Claude Code on all paid plans. Released September 22, 2026.
Claude Sonnet 5.5 — Faster and cheaper than Opus: everyday tasks, edits, documents and spreadsheets. Released September 28, 2026.
Claude Fable 5.1 — The most capable model, for complex multi-step tasks. Released September 1, 2026. Not available on Free; on Pro it's paid through usage credits; on Max it counts toward your plan's limit (up to 50% of the weekly limit).
Claude Haiku 4.5 — The fastest, cheapest model, for simple tasks. Retirement from the API no earlier than October 15, 2026.
Claude Sonnet 5 — The default model on Free and Pro since June 30, 2026 (per the announcement). Anthropic hasn't officially said whether Free has switched to Sonnet 5.5.

Plans

Free: $0 · Chat on the web and in the desktop and mobile apps, web search, file creation, memory, artifacts and skills.
Pro: $20/month or $17/month billed yearly ($200) · Higher limits, Claude Code, Cowork, Claude Design, scheduled tasks, Projects, more models.
Max: $100/month (5x) to $200/month (20x) · 5 or 20 times Pro's limits, priority at peak times.
Team: $25/month per seat ($20 billed yearly); Premium seat $125 ($100 billed yearly) · 2 to 150 people. Team data isn't used for training by default.
Enterprise: $20 per seat per month, billed yearly, plus usage at API rates · For organizations.

Notes

Training on your data: on Free, Pro and Max, Anthropic trains models on your chats only if the setting to help improve Claude is on. The toggle is at claude.ai/settings/data-privacy-controls. With it on, data is kept for up to 5 years; with it off, for 30 days.
Fable 5 and 5.1 by plan: on Max and a Team Premium seat, you can spend up to 50% of your weekly limit on Fable at no extra cost; on Pro and a Team Standard seat, Fable is outside the limit and paid through usage credits. The promotion that included Fable in the Pro and Team Standard limits ended July 19, 2026.
"Extra usage" was renamed "usage credits" (May 2026).

Sources: claude.com · platform.claude.com · support.claude.com · anthropic.com · anthropic.com · claude.com · code.claude.com · anthropic.com · support.claude.com · code.claude.com

ChatGPT

OpenAI · chatgpt.com

ChatGPT in the tool catalog

Models

GPT-6 Astra — OpenAI's most powerful model: announced September 3, 2026, then rolled out to paid plans. In regular chat it's on Pro; on Plus, only in the Work and Codex modes.
GPT-6.1 Sol — Close to Astra for less: long work with code and documents. Only in Work and Codex, not in regular chat. Not on Free or Go at launch.
GPT-6 Sol — Replaces GPT-5.5 for Codex on Plus and above (since September 22, 2026).
GPT-6 Luna — The most economical: extraction, summaries, simple tasks. For Free and Go in the desktop app, as it rolls out.
GPT-5.6 Sol / Terra / Luna — The regular chat models. Luna is the default on Free and Go, Sol on paid plans.
GPT-5.5 — Leaves ChatGPT, Work and Codex on October 14, 2026, on all plans; stays in the API.
gpt-image-2 — Built-in image generation in ChatGPT.

Plans

Free: $0 · Regular chat; Codex in the desktop app with GPT-6 Luna as it rolls out.
Go: $8/month · More messages, files and images than Free.
Plus: $20/month · Codex on the web, in the terminal, in IDEs and on iOS; GPT-6.1 Sol and GPT-6 Luna; Astra in Work and Codex; extra credits for purchase.
Pro: $100, $200 or $500/month · Three Pro tiers; Astra in regular chat; Astra Ultrafast mode on the $500 tier.
Business: $25/month per user ($20 billed yearly, 2 users minimum) · Business data isn't used for training by default.

Notes

Work mode (an agent for tasks) and Codex share one limit and one pool of credits.
To turn off training: Settings → Data controls → Improve the model for everyone. This also covers Codex on personal plans.
Sora as a standalone app shut down on April 26, 2026; the Sora API was turned off on September 24, 2026.

Sources: learn.chatgpt.com · learn.chatgpt.com · learn.chatgpt.com · help.openai.com · openai.com · help.openai.com · developers.openai.com · help.openai.com · openai.com

Gemini

Google · gemini.google.com

Gemini in the tool catalog

Models

Gemini 3.6 Flash — The app's main model for everyday tasks, available for free. Released July 21, 2026.
Gemini 3.1 Pro — A strong model for complex tasks; limited access on Free, more on paid plans.
Deep Think — A deep reasoning mode; the Ultra plan gets it first.
Nano Banana / Nano Banana Pro — Image generation and editing.
Gemini 4 Argon — Announced September 30, 2026, not yet publicly available; the paid API and Ultra get it first.

Plans

Free: $0 · 3.6 Flash, limited 3.1 Pro, images, Deep Research, Gemini Live, Canvas, 15 GB of storage.
Google AI Plus: $4.99/month (US) · 2x the limits, video, 200 Flow credits, 400 GB of storage.
Google AI Pro: $19.99/month (US) · 4x the limits, 1,000 Flow credits, Jules, the Gemini Spark agent (in select countries, 18+), 5 TB of storage.
Google AI Ultra: $99.99/month (5x Pro) or $199.99/month (20x) · Deep Think, the Gemini Spark agent (in select countries), Project Genie, 20 TB of storage and up.

Notes

Prices differ outside the US: the pricing page shows your local currency.
To stop your chats being used to improve Google's models: Settings → Activity → turn off Keep Activity. Temporary chats aren't used to improve the models.
On September 30, 2026, Google announced that Skills will replace Gems. Skills arrive in the Gemini app starting October 13, 2026, rolling out through mid-November. Gems move over to Skills automatically: personal accounts in November 2026, Workspace work accounts no earlier than March 1, 2027, education accounts no earlier than June 1, 2027.
NotebookLM has been called Gemini Notebook since July 16, 2026.
A Windows app has been available since September 10, 2026.

Sources: gemini.google · gemini.google · blog.google · blog.google · support.google.com · support.google.com · blog.google · gemini.google · support.google.com · workspaceupdates.googleblog.com

Microsoft Copilot

Microsoft · copilot.microsoft.com

Microsoft Copilot in the tool catalog

Models

Copilot — Microsoft doesn't say on its pricing page which models run inside.

Plans

Free: $0 · A limited free version; depends on available capacity.
Microsoft 365 Personal: $9.99/month or $99.99/year · Expanded Copilot chat in the Office apps.
Microsoft 365 Family: $12.99/month or $129.99/year · AI features for the subscription owner only.
Microsoft 365 Premium: $19.99/month or $199.99/year · The highest Copilot limits among the home plans, limited agent mode. Replaced Copilot Pro.
Microsoft 365 Pro: $99.99/month · The highest limits and agent mode.
Microsoft 365 Copilot Business: $18/month per user billed yearly (until December 31, 2026), $25.20 billed monthly · An add-on to a Microsoft 365 business plan; Copilot Chat is included in business plans at no extra cost.

Notes

Copilot Pro is no longer sold; support ended August 1, 2026.
To turn off training: your profile → Privacy → Training on conversation activity and Training on voice conversations.
Copilot isn't available in China (except Hong Kong) or in embargoed markets.

Sources: microsoft.com · support.microsoft.com · support.microsoft.com · microsoft.com · support.microsoft.com

Grok

SpaceXAI (formerly xAI) · grok.com

Grok in the tool catalog

Models

Grok 4.7 — The company's most powerful model (September 2026): code, chat, general tasks.
Grok 4.6 / 4.5 — Previous versions (August and July 2026).
Grok Imagine — Image and video generation.

Plans

Free: $0 · You can start using Grok for free.
SuperGrok: $30/month · Higher limits, image and video generation. Since June 2026, paid plans have one shared weekly limit across all Grok products.
SuperGrok Plus: $100/month · Much higher limits, 1080p video, priority at peak times.

Notes

SpaceX bought xAI in February 2026; the company now goes by SpaceXAI.
Access is no longer tied only to an X Premium subscription: there's grok.com and the apps.
You can turn off the use of your data for training in the Grok app settings and on grok.com.

Sources: docs.x.ai · docs.x.ai · docs.x.ai · x.ai · x.ai · x.ai · docs.x.ai

DeepSeek

DeepSeek · chat.deepseek.com

DeepSeek in the tool catalog

Models

DeepSeek-V4.1-Flash — The current main model (September 10, 2026): text, agents, image understanding. The weights are published on Hugging Face.
DeepSeek-V4-Pro — Since September 14, 2026, requests to it are routed to V4.1-Flash until V4.1-Pro comes out.

Plans

Chat on the website and in the app: no paid plans on the website · Only the API is paid (pay per token).

Notes

User data is stored and processed in the People's Republic of China (privacy policy, updated February 10, 2026).
To opt out of your data being used for training, send a request to privacy@deepseek.com.

Sources: deepseek.com · deepseek.com · api-docs.deepseek.com · cdn.deepseek.com · huggingface.co

Mistral Vibe (formerly Le Chat)

Mistral AI · chat.mistral.ai

Mistral Vibe (formerly Le Chat) in the tool catalog

Models

Vibe Chat / Vibe Work / Vibe Code — Three modes: regular chat, an agent for work tasks, and a developer mode (CLI, VS Code, web). The help center doesn't say which models run inside.

Plans

Free: $0 · Limited messages and web search.
Pro: $14.99/month · More messages, coding in the CLI and IDEs; $5.99/month for students.
Team: $24.99/month per user ($50/month minimum) · Up to 30 GB per user.

Notes

Le Chat was renamed Vibe on August 12, 2026; same address, same account.
On Free, your data is used for training by default. To turn it off: Privacy → turn off Allow your interactions to be used to train our models; in the app, Data & Account Controls → uncheck Enable data sharing. Vibe and the API have separate toggles.

Sources: mistral.ai · help.mistral.ai · help.mistral.ai

Perplexity

Perplexity · perplexity.ai

Perplexity in the tool catalog

Models

Models from several companies — On paid plans you can pick the model for each request; Perplexity's own search model is Sonar.

Plans

Free: $0 · Search with sources.
Pro: $20/month · Perplexity Computer, a choice of model for each request.
Max: $200/month · Everything in Pro with higher limits, 10,000 credits a month.

Notes

To turn off training: account settings → Preferences → AI data retention (called AI Data Usage in some versions of the interface). It applies only to data collected after you turn it off.

Sources: perplexity.ai · perplexity.ai · perplexity.ai · perplexity.ai · docs.perplexity.ai · perplexity.ai

Meta AI

Meta · meta.ai

Meta AI in the tool catalog

Models

Muse Spark — The model that has powered Meta AI since April–May 2026: in the Meta AI app, on meta.ai, and in WhatsApp, Instagram, Facebook and Messenger.
Muse (agent) — A separate personal agent, announced September 8, 2026; rolling out in the US (iOS, Android, muse.ai).

Plans

Meta AI: $0 · Built into WhatsApp, Instagram, Facebook and Messenger.
Muse Power: $20/month · Limited testing; not available in every country.
Muse Maximum: $100/month · Limited testing.

Notes

Incognito Chat is rolling out in WhatsApp and the Meta AI app over the coming months (announced May 13, 2026). According to Meta, not even Meta can read these chats, and messages disappear by default.
For how Meta uses data to train its models and whether you can opt out, see facebook.com/privacy/genai.

Sources: about.fb.com · about.fb.com · meta.com · about.fb.com · facebook.com

Claude Code today

Version

2.1.289 (03.10.2026)

What it is

Anthropic's coding agent: it reads your project, edits files and runs commands. It works in the terminal, in VS Code and JetBrains, in the Claude desktop app (the Code tab), in the browser (claude.ai/code) and on your phone.

Install

Where	How	Note
Terminal: macOS, Linux, WSL	curl -fsSL https://claude.ai/install.sh \| bash	The recommended option; it updates itself. Then cd into your project folder and run claude.
Windows PowerShell	irm https://claude.ai/install.ps1 \| iex	Git for Windows is recommended but not required: without it, Claude Code uses PowerShell.
Homebrew	brew install --cask claude-code	Doesn't update itself: run brew upgrade claude-code.
WinGet	winget install Anthropic.ClaudeCode	Doesn't update itself.
npm	npm install -g @anthropic-ai/claude-code	Still works and needs Node.js 22+, but the native installer is now the main option.
Desktop app	Download Claude for macOS or Windows, sign in, open the Code tab	No terminal needed. Requires a paid plan. Linux (Ubuntu, Debian) is in beta.
Browser and phone	claude.ai/code and the Claude app for iOS and Android	Tasks run in the cloud; you need a repository on GitHub.
VS Code, Cursor, JetBrains	The Claude Code extension from the extension marketplace	JetBrains also needs the CLI installed separately.

Requirements

macOS 13+, Windows 10 1809+, 4 GB RAM.

Plans

Plan	Access	Note
Free	No	Claude Code comes with paid plans.
Pro ($20/month)	Yes	Default model Opus 5.5; auto mode; routines; artifacts; computer use.
Max (from $100/month)	Yes	The same features with higher limits.
Team and Enterprise	Yes	Plus Code Review for GitHub and Claude Tag in Slack.
Anthropic Console (API key)	Yes, pay per token	No desktop app, cloud sessions, routines or artifacts.

Models

Name	Use
Opus 5.5	The default model on Pro, Max, Team, Enterprise and the API (since version 2.1.280). Default effort level: medium.
Sonnet 5.5	Faster and cheaper; alias sonnet. Requires version 2.1.284+.
Fable 5.1	For the longest, hardest tasks; you turn it on yourself (/model fable). On subscriptions it may draw on usage credits, and Claude Code asks for your consent first.
Haiku 4.5	Simple, quick tasks.
opusplan	Opus in plan mode, Sonnet for execution.

Features

Name	What it is
CLAUDE.md	A file with your project's rules; Claude reads it at the start of every session. Create it with /init. Claude also reads AGENTS.md; rules for specific file types go in .claude/rules/.
Auto memory	Claude records what it learns from your corrections and keeps it between sessions. Manage it with /memory.
Skills	A SKILL.md file with instructions; call it as /name, or Claude picks it up automatically from its description. Custom commands in .claude/commands/ have been merged into skills and keep working.
Subagents	Helpers with their own context that return a summary. They run in the background by default; create them by asking Claude or with files in .claude/agents/.
Hooks	Automatic actions on events: a script, an HTTP request, an MCP call, a prompt or a subagent. For example, formatting after every edit.
MCP	Connects outside services (Drive, Jira, Slack, databases). Your claude.ai connectors carry over when you sign in with a subscription.
Plugins	Bundles skills, hooks, subagents and MCP into one installable package; install them from a marketplace with /plugin.
Permission modes	Manual (setting value: default), acceptEdits, plan, auto, dontAsk, bypassPermissions. Switch with Shift+Tab. Since version 2.1.283 the default starting mode is auto: instead of asking you, a separate classifier model checks each action.
Plan mode	Claude studies the project and proposes a plan; it edits only after you approve. Enter it with Shift+Tab or /plan.
Routines	Cloud tasks that run on a schedule, from an API call or on GitHub events; they work even when your computer is off. Pro, Max, Team, Enterprise. Create them at claude.ai/code/routines or with /schedule. Status: research preview.
Desktop scheduled tasks and /loop	Schedules on your own computer with access to local files; /loop repeats a prompt inside an open session.
Artifacts	A live web page on claude.ai built from a session's results: a dashboard, a report, a comparison. Private by default. Pro, Max, Team, Enterprise.
Background agents and agent view	The claude agents command opens a view of all background sessions; /background sends the current session to the background. Research preview.
Agent teams	Several sessions sharing one task list, led by a lead session. Experimental, off by default.
Dynamic workflows	A script that runs many subagents and double-checks their results; view them with /workflows. Built for audits and large migrations.
Projects	One long conversation in claude.ai/code or the app, where Claude starts parallel sessions on its own. Public beta on Pro and Max.
Remote Control and phone	Continue a local session from your phone or browser; claude --teleport pulls a cloud session into your terminal.

Changes since May 2026

What it is	Date
New models: Opus 4.8 (May 28), Fable 5 (June 9), Sonnet 5 (June 30), Opus 5 (July), Fable 5.1 (September 1), Opus 5.5 (September 22), Sonnet 5.5 (September 28).	2026-05..09
The default model on all subscriptions is now Opus 5.5 (it used to be Sonnet on Pro).	2026-09 (v2.1.280)
Auto mode came to Pro (May), then became the default starting mode (v2.1.283, September). The default mode is called Manual in the interface.	2026-05..09
"Extra usage" was renamed "usage credits", and the /extra-usage command became /usage-credits (the old name still works).	May 2026
New: agent view (claude agents), /goal, dynamic workflows, /cd, artifacts, Claude in Chrome out of beta, subagents in the background by default, fork mode by default, cross-session messaging, /design, /slides, /code-review with levels, /doctor as a full checkup (alias /checkup), /skill-doctor, Ultracode in /effort, mods (plugins with an interface).	2026-05..09
Desktop app: Linux in beta, a built-in browser, an iOS simulator, and the /desktop command to continue a terminal session in the app.	2026-06..08
Claude Code in Slack for Team and Enterprise is being replaced by Claude Tag.	06/23/2026
Vertex AI is now called Google Cloud's Agent Platform in the docs.	2026
Windsurf is renamed Devin Desktop in the /ide and /terminal-setup commands.	June 2026

Removed

What it is	Date	Instead
/ultraplan and the ultraplan keyword	August 2026	plan mode or Claude Code in the cloud
The /agents wizard	v2.1.198	ask Claude to create a subagent, or edit .claude/agents/; /agents now only shows a hint
/review as a separate command	v2.1.223	/review is now an alias for /code-review
/vim	v2.1.92	/config → Editor mode
/pr-comments	v2.1.91	ask Claude to show the PR comments
The TaskOutput tool	September 2026	Claude reads the background task's output file with Read
Fast mode on Opus 4.6	May 2026	fast mode on Opus 5.5 ($8/$40 per 1M tokens)

Notes

The /import command brings over settings from OpenAI Codex, Gemini CLI and Cursor.

Plans and prices

Plans and prices
Service	What it is	Price	Source
n8n	Community Edition (self-hosted)	Free	n8n.io
n8n	Cloud Starter: 2,500 executions a month	€20/month billed yearly	n8n.io
n8n	Cloud Pro: 10,000 executions a month	€50/month billed yearly	n8n.io
Zapier	Free: 100 tasks a month, two-step Zaps only	$0	zapier.com
Zapier	Professional: from 750 tasks a month, multi-step Zaps, webhooks	from $19.99/month billed yearly or $29.99 billed monthly	zapier.com
Zapier	Team: up to 25 users	from $69/month billed yearly or $103.50 billed monthly	zapier.com
Make	Free: 1,000 credits a month, 2 active scenarios	$0	make.com
Make	Core / Pro / Teams: 10,000 credits a month each (billed in credits, not operations)	$9 / $16 / $29 a month billed monthly; 15% or more off billed yearly	make.com
ElevenLabs	Free: 10,000 credits, no commercial license	$0	elevenlabs.io
ElevenLabs	Starter / Creator / Pro (with a commercial license)	$6 / $22 (first month $11) / $99 a month	elevenlabs.io
Suno	Free: 50 credits a day, v6-mini model, no commercial rights	$0	suno.com
Suno	Pro: 2,500 credits, commercial rights, v6 models	$8/month (the page says yearly billing saves $24 a year)	suno.com
Suno	Premier: 10,000 credits	$24/month (yearly billing saves $72 a year)	suno.com
Vapi	Voice agent platform: Vapi charges per minute; speech recognition, the LLM, the voice and phone service are billed separately at each provider's rates	$0.05/min + providers; $5 in free credits to start	vapi.ai
Cloudflare Workers	Free: 100,000 requests a day, 10 ms of CPU per request	$0	developers.cloudflare.com
Cloudflare Workers	Paid: 10 million requests a month included, then $0.30 per million	from $5/month	developers.cloudflare.com
Cloudflare Pages	Free: 500 builds a month, up to 20,000 files per site; Pages Functions are billed as Workers	$0	developers.cloudflare.com
Stripe	Card processing fee, US	2.9% + 30¢; +1.5% for international cards; +1% if currency conversion is needed	stripe.com
Stripe	Card processing fee, Canada (an example: the rate depends on the country)	2.9% + CA$0.30; +0.8% for international cards; +2% for currency conversion	stripe.com
Vercel	Hobby: personal, non-commercial projects only	$0	vercel.com
Vercel	Pro: per developer, includes $20 of usage credit	$20/month	vercel.com
Supabase	Free: 2 active projects, 500 MB database, 50,000 MAU (monthly active users), paused after a week without activity	$0	supabase.com
Supabase	Pro: 100,000 MAU, 8 GB per project, $10 in compute credits	$25/month	supabase.com
Anthropic API	Claude Fable 5.1, per 1M tokens (input / output)	$10 / $50	platform.claude.com
Anthropic API	Claude Opus 5.5, per 1M tokens	$4 / $20	platform.claude.com
Anthropic API	Claude Sonnet 5.5, per 1M tokens	$2 / $10	platform.claude.com
Anthropic API	Claude Haiku 4.5, per 1M tokens	$1 / $5	platform.claude.com
Anthropic API	Discounts and surcharges	Batch API −50%; cache reads at 10% of the input price (Opus 5.5: 5%, Fable 5.1: 2.5%); web search $10 per 1,000 searches; 1M-token context at no extra charge	platform.claude.com
OpenAI API	gpt-6-astra, per 1M tokens (short context)	$10 / $50	developers.openai.com
OpenAI API	gpt-6.1-sol and gpt-6-sol, per 1M tokens	$2 / $10	developers.openai.com
OpenAI API	gpt-6-luna, per 1M tokens	$0.10 / $0.50	developers.openai.com
OpenAI API	gpt-5.6-sol / gpt-5.6-terra / gpt-5.6-luna, per 1M tokens	$4 / $20; $2 / $12; $0.20 / $1.20	developers.openai.com
OpenAI API	Discounts	Batch and Flex −50%; long context costs more	developers.openai.com
Gemini API	gemini-3.8-flash, per 1M tokens	$0.75 / $3.75 through December 31, 2026; $1.50 / $7.50 from January 1, 2027; free tier available	ai.google.dev
Gemini API	gemini-3.1-pro-preview, per 1M tokens (prompts up to 200k)	$2 / $12; no free tier	ai.google.dev
DeepSeek API	deepseek-flash (V4.1-Flash), per 1M tokens, cache miss / output	$0.30 / $1.20 at peak hours; half that the rest of the time	api-docs.deepseek.com
SpaceXAI API (Grok)	grok-4.7, per 1M tokens	$2 / $6 ($4 / $12 above 200k tokens)	docs.x.ai

Renamed and closed

Renamed

Renamed
Before	What's new	Date	Source
Le Chat (Mistral)	Mistral Vibe (modes: Vibe Chat, Vibe Work, Vibe Code)	08/12/2026	help.mistral.ai
NotebookLM	Gemini Notebook (notebook.google)	07/16/2026	blog.google
Gems (Gemini)	Skills in Gemini; Gems move over automatically (work accounts no earlier than March 1, 2027)	November 2026	support.google.com
Windsurf	Devin Desktop (windsurf.com redirects to devin.ai/desktop)	06/02/2026	devin.ai
Freepik	Magnific (freepik.com redirects to magnific.com)	04/28/2026	magnific.com
xAI	SpaceXAI (after SpaceX bought xAI)	02/02/2026	x.ai
Copilot Pro	Microsoft 365 Premium ($19.99/month)	08/01/2026	support.microsoft.com
Claude Code in Slack (Team and Enterprise)	Claude Tag	06/23/2026	code.claude.com
Extra usage, /extra-usage (Claude)	Usage credits, /usage-credits	May 2026	code.claude.com
/review (Claude Code)	an alias for /code-review	v2.1.223	code.claude.com
Vertex AI (in the Claude Code docs)	Google Cloud's Agent Platform	—	code.claude.com
Captions (the company)	Mirage (the Captions app keeps its name)	2025	captions.ai
v0.dev	v0.app	≤ 2026-10-04	v0.app
runwayml.com	runway.com	≤ 2026-10-04	runway.com
klingai.com	kling.ai	≤ 2026-10-04	kling.ai
suno.ai	suno.com	≤ 2026-10-04	suno.com
notion.so	notion.com	≤ 2026-10-04	notion.com
fathom.video	fathom.ai	≤ 2026-10-04	fathom.ai
blackforestlabs.ai	bfl.ai	≤ 2026-10-04	bfl.ai
usegalileo.ai (Galileo AI)	stitch.withgoogle.com (Google Stitch)	May 2025	x.com

Closed

Closed
Name	Date	Note	Source
Sora (website and app)	04/26/2026	OpenAI's help article, linked in the Source column, explains how to save your videos.	help.openai.com
Sora 2 API and Videos API	09/24/2026	No replacement in the OpenAI API.	developers.openai.com
DALL-E 2 and DALL-E 3 in the API	05/12/2026	Replacement: gpt-image-2. In ChatGPT, images are made by gpt-image-2.	developers.openai.com
OpenAI Assistants API	08/26/2026	Replacement: Responses API and Conversations API.	developers.openai.com
OpenAI Agent Builder	11/30/2026	Scheduled. Replacement: Agents SDK or Workspace Agents in ChatGPT. ChatKit stays.	developers.openai.com
OpenAI Evals (platform)	11/30/2026	Scheduled; read-only from October 31, 2026.	developers.openai.com
GPT-5.5 in ChatGPT, Work and Codex	10/14/2026	Scheduled on all plans; stays in the API.	learn.chatgpt.com
GPT-5.4 and GPT-5.4-mini in Codex with ChatGPT sign-in	08/31/2026	Replacement: GPT-6 Sol and GPT-6 Luna.	learn.chatgpt.com
GPT-5.3-Codex-Spark	09/14/2026	Removed from the app, CLI and IDE.	learn.chatgpt.com
Copilot Pro	08/01/2026	Sales stopped earlier; support ended August 1, 2026. Replacement: Microsoft 365 Premium.	support.microsoft.com
Gems in Gemini	November 2026	Announced September 30, 2026: Skills replace Gems, and Gems move over automatically. Personal accounts: November 2026; Workspace work accounts: no earlier than March 1, 2027; education accounts: no earlier than June 1, 2027.	support.google.com
DeepSeek-V4-Pro	09/14/2026	Requests are routed to V4.1-Flash until V4.1-Pro comes out.	deepseek.com
Claude Opus 4, Opus 4.1, Sonnet 4, Haiku 3.5 in the Claude API	≤ 2026-10	Retired from the Claude API; some remain available on Bedrock and Google Cloud.	platform.claude.com
Claude Haiku 4.5	≥ 2026-10-15	The earliest possible date for retirement from the API; watch the model deprecations page.	platform.claude.com
Ultraplan in Claude Code (/ultraplan)	August 2026	Use plan mode or Claude Code in the cloud.	code.claude.com
The /agents wizard in Claude Code	v2.1.198	Create subagents by asking Claude or with files in .claude/agents/.	code.claude.com