generated: '2026-10-18'
method: searched
generator: extract-docs-artifacts.py (local)
source: https://llmstatus.ai/deprecations
sources:
- https://llmstatus.ai/deprecations
- https://llmstatus.ai/changelog
entries:
- breaking: false
  highlights:
  - 'See exactly what will text or email you New Delivery view: every enabled rule grouped by the channel it sends to — SMS first, then email,
    chat, in-app — so "what will text me?" has a one-glance answer.'
  - On the web Alerts page, in the TUI Alerts tab (← → Delivery; Enter edits the rule), and on the command line as mm alerts channels.
  - 'Tuning what reaches you is per rule: scope (only models you use · chosen providers or families · everything), events, and channels.'
  - Edit a rule from any of those views, or mm alerts edit <rule> --scope mine --channels email.
  date: '2026-10-10'
- breaking: false
  highlights:
  - One Scan tab, straight to Inventory The Here and Scan tabs are now one tab, Scan.
  - It scans the folder on open; signed in, u syncs the whole result to your cloud inventory in one step (adds what's here, removes what's gone)
    and lands you on Inventory with a banner saying exactly what changed ("synced 8 · 3 new · 5 updated · removed 2 no longer in the code").
  - 'Tabs renumbered: 1 Scan · 2 What''s New · 3 Inventory · 4 Add · 5 Alerts · 6 Account · 7 Sources.'
  - Donkey Kong ( P ) is still there while a scan runs.
  date: '2026-10-10'
- breaking: false
  highlights:
  - 'When a newer model in a family you use comes out, you get a message about your situation, not a generic announcement: "Claude Haiku 5.5 is
    out — you use Claude Haiku 4.5 in 2 places", with the call sites, price per 1M tokens side by side ("$0.10 / $0.50 vs $1 / $5 — 10× cheaper
    on input"), context window, max output, and the old model''s retirement date.'
  - Only accounts using the older model get it.
  - The registry now carries list prices (from models.dev) and a tier-aware family for every model, shown on model pages and in the signed snapshot.
  - 'Rules can listen to specific model families, not just whole providers: "Anthropic, plus anything in GPT mini".'
  date: '2026-10-10'
- breaking: false
  highlights:
  - Your inventory follows the code Scanning a repo now removes tracked references that are no longer in the code.
  - Before, an upload only ever added, so after you migrated off a retiring model your inventory kept the old references and kept sending retirement
    alerts for them.
  - mm scan , the TUI's upload ( u ), and mm ci --report all reconcile; each one tells you what it removed.
  - Retirement reminder emails now say what to do if you've already migrated.
  date: '2026-09-27'
- breaking: false
  highlights:
  - Honest counts from mm prompt mm prompt's summary line (and the TUI p toasts) now report the same model/reference counts as the prompt they
    just generated — previously duplicate references at the same spot were counted before deduping, so the summary could say "3 references" while
    the prompt said "1 in total".
  - The prompt's "checked against the LLM Status registry on …" date now uses your local calendar day instead of UTC, so an evening run no longer
    stamps tomorrow's date.
  - When copying to the clipboard fails, the TUI now says "couldn't copy — prompt saved to <file>" instead of blaming a missing clipboard tool
    that may actually be installed.
  - 'Help-text touch-ups: --model is documented for mm prompt too, and mm sources separates --db and --sql-table correctly.'
  date: '2026-08-30'
- breaking: false
  highlights:
  - 'Hand the fixes to your AI coding agent New mm prompt [dir] : prints one self-contained prompt describing every retired/retiring model reference
    in your repo (file:line, the exact string in code, the chain-resolved replacement) plus rewrite rules.'
  - 'Paste it into Claude Code, Cursor, or any coding agent — or pipe it straight to your clipboard: mm prompt | pbcopy.'
  - Unlike mm fix (a mechanical in-place rewrite), an agent can also re-pin dated ids like -20250514 , follow config indirection, and adjust parameters
    the new model needs — the prompt spells out those rules for it.
  - 'TUI: press p on the Here tab or the Inventory tab to copy the same prompt to your clipboard (with a file fallback when no clipboard tool
    exists).'
  date: '2026-08-22'
- breaking: false
  highlights:
  - 'Hardening pass: 40 fixes from an adversarial audit mm fix now writes the provider’s real model id ( gpt-5.5 ), not the registry’s dash-form
    slug ( gpt-5-5 ) — 59 of 85 replacement targets were affected.'
  - It also fixes everything the scanner detects (Bedrock ARNs, dated variants, mixed case), preserves your file’s line endings, and refuses to
    touch files that aren’t valid UTF-8.
  - 'The CI gate can’t pass by accident anymore: --fail-on=retiring (equals-style flags) now parses, a typo’d --sources or a path that doesn’t
    exist is a hard error instead of a silent green build, and withdrawn models fail the gate like retired ones.'
  - mm scan --dry-run is fully local — no account needed, nothing sent anywhere.
  date: '2026-07-07'
- breaking: false
  highlights:
  - Search the What's New feed The What's New tab now has / search, like Scan and Inventory.
  - Filter the registry feed, alerts, fixes, and releases by typing — e.g. deprecated to see just the deprecations, or a model or provider name
    to jump to a specific change.
  - Press / to start filtering, esc to clear; the count of matches shows as you type.
  date: '2026-06-16'
- breaking: false
  highlights:
  - Command polish + safer billing mm upgrade --lifetime starts the one-time Lifetime checkout from the CLI (plain mm upgrade is still the $5/yr
    Pro plan).
  - mm <command> --help now prints per-command usage and flags; an unknown command exits non-zero instead of silently printing the help.
  - mm clear requires an explicit --yes / --force — it will no longer wipe your cloud inventory on --ci alone.
  - mm logout no longer claims it removed a key when you weren't signed in, and warns if an env-var key still authentic.
  date: '2026-06-14'
current_version: v 0.1.93
note: Versioning scheme follows the pattern "v X.Y.Z" as shown in the changelog entries.
docs: https://llmstatus.ai/deprecations