Every pricing change, policy shift, and correction — logged publicly. Corrections show the original claim (strikethrough) and the fix. Nothing is silently edited.
Note: live DB feed coming soon. Current entries are manually curated.
At Google I/O 2026, Google launched four Gemini features. (1) Daily Brief — personalized morning digest from Gmail, Calendar, and Tasks — rolling out now to Google AI subscribers. (2) Gemini Spark — 24/7 background cloud agent with standing Gmail inbox access, runs when phone is locked, custom workflows — arriving next week for Google AI Ultra subscribers only. Data governance scope and opt-out paths not published at launch. (3) Gemini Omni — AI video model combining Gemini and generative media models — for Google AI subscribers via Google Flow and YouTube Shorts. (4) Neural Expressive UI — free for all users. Access note: Spark requires Google AI Ultra (~$249.99/month), the highest consumer tier — above ChatGPT Pro ($200/month) and Claude Max ($100/month). Google AI plan (~$19.99/month) required for Daily Brief and Omni.
Three Claude changes between May 6–7, 2026. (1) Claude Code five-hour rate limit doubled for Pro, Max, Team, and Enterprise — peak-hour throttling removed for Pro/Max. Free tier unchanged. Powered by new SpaceX compute deal (220,000+ NVIDIA GPUs). (2) Claude Word, Excel, and PowerPoint add-ins went GA; Outlook entered public beta. No extra charge beyond existing Pro subscription ($20/mo), vs Microsoft Copilot's $30/mo. (3) "Dreaming" launched for Claude Managed Agents: a scheduled background process that scans past agent sessions and reorganizes memory between tasks. Not model training — but a background read of session history with no opt-out surfaced at launch. Users relying on Claude Managed Agents should verify whether session data review is acceptable under their use case.
OpenAI launched Codex mobile access on May 14, 2026 for iOS and Android, available to all ChatGPT plans including the free tier. The mobile interface is a remote control for Codex sessions running on desktop or cloud — not a standalone on-device coding agent. Users can monitor tasks, review outputs, and dispatch instructions from mobile. By contrast, Claude Code requires a paid Claude subscription (minimum $20/month Pro). Free-tier users are subject to OpenAI's standard data retention policies; code processed via Codex is handled on OpenAI servers under standard terms.
OpenAI expanded ad targeting on Free and Go tiers to use conversation topics, past chat history, and past ad interactions within ChatGPT. Geographic expansion: US rollout extended to Canada, Australia, and New Zealand. Plus, Pro, Business, Enterprise, and Education plans remain ad-free. OpenAI has not confirmed whether conversation data is shared with third-party ad networks.
OpenAI replaced GPT-5.3 Instant with GPT-5.5 Instant as the default ChatGPT model across all tiers — Free, Plus, Pro, Team, Business — with no in-product notification. GPT-5.5 Instant also replaced GPT-5.3 Instant as the chat-latest API endpoint, silently migrating developers who relied on that alias. GPT-5.3 Instant remains selectable in model settings for three months before retirement. Note: GPT-5.5 Instant is a separate model from GPT-5.5 (the premium API model launched April 23 at $5/$30 per million tokens).
Google launched chat history import at gemini.google.com/import (March 26, global settings rollout April 20). Users can upload ZIP exports from ChatGPT or Claude — up to 5 files/day, 5GB each. Gemini Apps Activity is on by default: imported conversations may be used for model training unless manually disabled at myactivity.google.com → Gemini Apps Activity. Feature is unavailable in EEA, UK, and Switzerland. Opt-out is not surfaced inside the import flow.
OpenAI released GPT-5.5 on April 23, 2026, exactly seven days after Claude Opus 4.7. API pricing doubled from $2.50/$15 to $5/$30 per million input/output tokens. OpenAI claims the "effective cost" is only 20% higher due to token efficiency gains — meaning roughly 60% of efficiency improvement was captured as margin rather than passed to users. Claude Opus 4.7 (released April 16) kept API pricing flat at $5/$25 for a comparable capability improvement. GPT-5.5 leads on agentic workflows (82.7% Terminal-Bench 2.0) and long-context reasoning (74.0% MRCR v2, up from 36.6%). Claude Opus 4.7 leads on coding (64.3% vs 58.6% SWE-bench Pro).
GitHub removed all Claude Opus models from Pro ($10/mo) effective April 20, 2026. Pro+ ($40/mo) lost Opus 4.5 and 4.6; only Opus 4.7 remains at a 7.5x request multiplier (previously 3x), cutting effective monthly quota by 60%. New signups frozen for Pro, Pro+, and Student plans. Zero advance notice given — changes effective within days of the April 17 announcement. Self-service refunds available until May 20, 2026 (Settings → Billing → Cancel and refund). Potential violation of Quebec Consumer Protection Act Section 11.2 (requires 30-day written notice for unilateral contract changes).
Verified current pricing for Perplexity Pro tier. No change from previous snapshot. Context window confirmed at 127K tokens.
Mistral Le Chat Pro ($14.99/mo) added as tracked tool. Pricing sourced directly from mistral.ai/pricing. Context window: 128K tokens.
Previously published: 128K tokens. Corrected to 131K tokens after re-verification against xAI documentation. Original error was due to rounded figure in a third-party source.
Methodology — how changes are detected and verified
Manifesto — why this changelog exists