LLM Status — AI model deprecation tracker & CLI checker
mm status — ~/code/api-service
$ mm status●387 ◐41 ▲20 ✕79
LLM Status — registry live · 15 providers · 527 models<br>Scanned ~/code/api-service · 6 models tracked
◐deprecatingopenai/gpt-5-2025-08-07retires 2026-12-11in 4mo
▲retiringamazon/amazon.nova-premier-v1:0retires 2026-09-14in 25d
●okanthropic/claude-opus-4-7retires 2027-04-16in 8mo
◐deprecatingopenai/o3-pro-2025-06-10retires 2026-12-11in 4mo
✕retiredmicrosoft/gpt-5.3-chatretires 2026-08-1010d ago
▲retiringopenai/gpt-3.5-turbo-instructretires 2026-09-28in 39d
⚠ 140 of 527 registry models are deprecated, retiring, or gone. Run mm for your repo's list.
↑↓ navspace selectq quit
These retirement dates are live from the registry. Repo path and scan counts are a sample.
Setup
Install it, then run mm.
Install<br>Installs the mm command — a self-contained binary, no Node needed (60–120 MB, auto-detects your platform). Auto-updates in the background, signature-verified; MM_NO_AUTO_UPDATE=1 disables it.<br>$ curl -fsSL https://llmstatus.ai/install.sh | bashCopied!<br>or brew install randomartifact/tap/modelstatus-cli · npm i -g @modelstatus/cli · try once with npx @modelstatus/cli status
Open the dashboard<br>Run mm for the full terminal dashboard: it scans the current repo on launch, then you arrow through every model it calls, browse the registry, and see what's new — live.<br>$ mmCopied!
Or just the check<br>Need it scriptable? mm status lists every model the repo calls, soonest retirement first. mm ci exits non-zero when something's on a clock — that's the one for pre-commit hooks and CI. No account.<br>$ mm ci . --fail-on retiringCopied!
Both pull the signed registry from cdn.llmstatus.ai once, then run offline. No account. Anonymous usage counts only — MM_NO_ANALYTICS=1 turns them off.
CLI
Run mm for the full dashboard.
Type mm to open it. Arrow through every model the repo calls, press f to fix one with a diff preview, and watch the ●◐▲✕ counts update as you go. The Here and What's New tabs read straight off the signed registry, no account, works offline.
$ mm
A real recording — mm 0.1.87: the fix flow, then the live registry feed with / search.
mm — ~/code/api-service
Fix
mm fix swaps in the replacements.
Finding dying models is half the job. One command rewrites them to the registry's replacement, in place.
mm — ~/code/api-service
Only the id text changes. Quotes, formatting, the rest of the line: byte-identical.
Boundary-safe: gpt-4 can't rewrite inside gpt-4o or gpt-4-tuned.
Chains through dying replacements to the first live model.
--dry-run previews, --json for tooling. In the TUI, press f.
Pull request · Swap retiring model ids for replacements llm-status[bot]
workers/classify.py
@@ line 88 @@
- model = "gpt-4o"
+ model = "gpt-5-1"
The GitHub App does the same on pull requests: a failing check grows an Open fix PR button, and one click opens this — the diff in the body, the commit verified, re-checked green by the same scan.
A real recording — mm 0.1.87 on a sample repo; dates and replacements come from the live registry.
Registry
Retiring soon
Live from the registry — it refreshes every 6 hours from models.dev, OpenRouter, and provider APIs.
On a clock
Models with an announced shutoff date. Click one for the full lifecycle.
Amazon Nova Premier<br>Amazon<br>2026-09-14<br>in 25d
GPT-3.5 Turbo Instruct<br>OpenAI<br>2026-09-28<br>in 39d
Claude Sonnet 4.5<br>Anthropic<br>2026-09-29<br>in 40d
Amazon Nova Canvas<br>Amazon<br>2026-09-30<br>in 41d
GPT-4o<br>OpenAI<br>2026-10-01<br>in 42d
Nano Banana<br>Google<br>2026-10-02<br>in 43d
Claude Haiku 4.5<br>Anthropic<br>2026-10-15<br>in 56d
Gemini 2.5 Flash<br>Google<br>2026-10-16<br>in 57d
Just released
Newest additions to the registry.
GG
Gemini 3.7 Flash 7d ago
Google · gemini-flash · 1049k ctx
XA
Grok 4.6 8d ago
xAI · grok · 500k ctx
AM
Grok 4.6 8d ago
Amazon · grok · 500k ctx
XA
Grok Imagine Image 2.0 13d ago
xAI · grok · 8k ctx
ME
Muse Spark 1.2 Contributor 15d ago
Meta · muse · 1049k ctx
No AI provider ships a model-lifecycle feed. We do —<br>open the radar on your phone, or subscribe:<br>RSS ·<br>JSON ·<br>retirements only
GitHub
Check it on every pull request.
Add a model-lifecycle check to every PR. Two options, and the difference is whether your source code leaves your runners.
GitHub Action
Runs in your own CI; your source never leaves your runners, only model IDs go out. Inline PR annotations and a pass/fail gate.
Free check · Pro adds cloud drift + alerts
$ npx @modelstatus/cli ci . --fail-on retiring
GitHub App Pro
One click, no workflow YAML or secrets to manage. We scan each PR on our servers (your code is sent to us; use the CI option if that's a dealbreaker) and post a Check Run plus a review comment. A button on the check opens a fix PR that swaps in the replacements.
Server-side · included with Pro
Connect your repos →
Read the CI docs →
Pricing
The CLI is free. Cloud is $5 a year.
Plans cover the cloud inventory and alerts. The CLI itself is always...