Cut your AI bill in half. Use your existing subscriptions. Every leading model inside Codex and Claude Code, and the right one for every task. Beta for macOS.
Joined October 2026
- Tweets29
- Following43
- Followers128
- Likes179
AutoRouter retweeted
I’ve been using @Auto_Router for a week+ now and haven’t run out of usage yet whereas I used to consistently bump into limits. The quality doesn’t seem to be down at all.
Plus, it’s elegantly designed and seamless to use. Highly recommend.
most of what you send fable or astra doesn't need fable or astra
renames, summaries, commit messages, small fixes. lighter models do those just as well
AutoRouter sends the hard stuff up and the rest down, so your subs go way further
haiku 5.5 runs a claude code task for $0.0044
opus runs an average task for $0.110 — 25× more
all 45 runs passed. on these small tasks, the cheap model passed every check
haiku's tokens are 40× cheaper than opus but its tasks only 25× — it takes more turns to get there
on the refactor: 8.7 turns vs opus's 4.3. wall time stayed within 9 seconds across all three models though
claude code's default is opus on pro, max & the api
for small tasks with a check you trust, run the smallest model that passes:
claude -p --model haiku "commit the staged changes"
keep opus for work only it can do
learn more: autorouter.com/blog/claude-c…
AutoRouter retweeted
the data is starting to come in from @Auto_Router's first week in the wild.
nothing too surprising here (Opus 5.5 is indeed OP on price/performance) — but I expect this to shift as Gemini 4 Carbon goes GA
give your agent a fast model for the small stuff
easy changes, renamed, and commit messages go to Claude Haiku 5.5 or GPT-6 Luna in AutoRouter
AutoRouter retweeted
There’s no universally “best” routing policy. The right model depends on the task—and your taste.
Opus 5.5 is my daily driver, but I don’t like its copywriting voice. For copy, I prefer GPT 6.1 Sol.
System One models like @typesafeai's Jev let you teach @Auto_Router those preferences in a few clicks.
How AutoRouter picks a model for each turn, in order:
1. Your rules
2. How hard the turn is
3. Only models that clear the quality bar for it
4. Your ChatGPT & Claude plans first for hard work
5. The cheapest capable model for everything else
Every turn shows the model, what it cost and why. It sometimes misjudges difficulty. Rate the pick and we retune.
AutoRouter retweeted
LiteLLM’s release pipeline was severely compromised two days ago as part of the mounting “vulnpocalypse.”
Having sat in the CTO seat, I’ve learned this lesson the hard way: when you self-host critical operational infrastructure, you own every incident, every patch, and every 2 a.m. wake-up call. That bill always comes due.
It’s one of the reasons we built @Auto_Router: a turnkey, secure, managed way for enterprises to govern their AI spend.
If the past few days have left you scrambling and you want a simpler, safer path, DM me. I’ll jump on a call and happily “Collison install” @Auto_Router for you.
Meet AutoRouter, the most intelligent way to pick models
You’re wasting tokens selecting frontier models for every prompt, and sometimes for worse performance
AutoRouter handles the decision layer:
- Understands the prompt request
- Picks the best model fit for the task
- Saves you money at every turn
- Plugs into your current setup
Works with your plans & every leading model inside Codex/Claude Code
Beta is live: autorouter.com
AutoRouter retweeted
Like clockwork the model wars has a new contender!
hello world
Meet AutoRouter, the most intelligent way to pick models
You’re wasting tokens selecting frontier models for every prompt, and sometimes for worse performance
AutoRouter handles the decision layer:
- Understands the prompt request
- Picks the best model fit for the task
- Saves you money at every turn
- Plugs into your current setup
Works with your plans & every leading model inside Codex/Claude Code
Beta is live: autorouter.com
AutoRouter retweeted
Meet AutoRouter, the most intelligent way to pick models
You’re wasting tokens selecting frontier models for every prompt, and sometimes for worse performance
AutoRouter handles the decision layer:
- Understands the prompt request
- Picks the best model fit for the task
- Saves you money at every turn
- Plugs into your current setup
Works with your plans & every leading model inside Codex/Claude Code
Beta is live: autorouter.com
AutoRouter retweeted
we're live in public beta today. try it at autorouter.com View article
AutoRouter retweeted
today we’re releasing AutoRouter for public alpha
run opus 5.5 or any other open source models inside codex or auto route by complexity to save big on frontier subscriptions
it attaches to codex & adds new models directly in the app’s picker so you’re not switching harnesses once your subscription runs out
rolling people off the waitlist today & tomorrow → autorouter.com