Pinned Tweet
Since all of our users seem to love DeepSeek V4.1 Flash 1017, we will be increasing our limits on our plans.
$10 Pro plan now gets extra $50($125 total) of usage on DeepSeek and the $30 Ultra Plan now gets extra $150($375 total) through November 30
Sign Up Now: Perchai.app
We are extending our free model promo till Nov 20 on these models so you can continue testing out our app:
1) DeepSeek v4.1 Flash (1017)
2) GLM 5.3 Flash
3) Qwen 3.8 Max
Try it out perchai.app
Let us know if you have any feedback or would like us to add anything!
Benchmarks officially out now for Hy4!
Kimi K3.1 Pro level intelligence at barely tenth of the price.
Try it out on Perch for 7.5x usage! perchai.app
👋Hy4 is here, new with vision.
Firmly among the leading open models.
More productive everywhere: finance, frontend, game, officework, research…
Built responsibly, with cyber/biosecurity in mind.
Try it out and let us know!
🔗:hy.tencent.ai/research/hy4?l…
Hy4 from Tencent landed a few hours ago. No benchmarks yet.
We put it on Perch anyway, so you can be the benchmark.
1M context · images · tools · effort from off to high
Pro + Ultra. Web, desktop, CLI: /model hy4
Get 7.5x credit for Hy4: perchai.app/pricing
Update: Stealth Model Crouton Alpha is now also available for Free on Perch AI.
Built for coding, agent work and high-volume tool use. 1M context, reads images.
Sign up at: perchai.app
Let us know what model y'all think it is!
We are extending our free model promo till Nov 20 on these models so you can continue testing out our app:
1) DeepSeek v4.1 Flash (1017)
2) GLM 5.3 Flash
3) Qwen 3.8 Max
Try it out perchai.app
Let us know if you have any feedback or would like us to add anything!
Why doesn't Perch Offer Model API Access?
One of the most common questions and request we get is if we offer or will offer API access? The short answer is No, we won't and that's by design.
The Longer Answer is: Perch for me is a passion project. It's something I built because I enjoy building it. The principle Perch AI is built on is that an economical token plan shouldn't have to mean subpar harness.
Users who have been here since Day 1 know this wasn't always a smooth ride. We started with a buggy app just on Mac in late June. Now we are available on every platform, plus CLI and Cloud Agents.
We could probably 10x our volume and revenue overnight if we offered API access but that's not what I want. I want to give y'all the best agentic app experience you can have free of cost which is why Perch outside of our inference doesn't cost a penny to run for people. We offer BYOK, BYOLM and soon the ability to run other CLI agents within our own GUI
Most people think we are also chasing token numbers like most our competitors, we are not. We'd rather provide 10B high quality token usage than offer 1T shitty ones.
Many people here think we are this highly funded massive team. We are not, I built Perch solo by myself and turned it into the entity it is today. This is why Perch is and will always be unique to most harnesses you will use.
If there's one thing I can promise y'all is that at Perch, you'll always be heard. Our current users can vouch for it. I am available to listen to your concerns via comments on X, messages on Discord or our email.
I want to thank you all for your continued support of Perch. I appreciate criticism as much as I do praise so please let us know how we can improve your experience going forward!
Since all of our users seem to love DeepSeek V4.1 Flash 1017, we will be increasing our limits on our plans.
$10 Pro plan now gets extra $50($125 total) of usage on DeepSeek and the $30 Ultra Plan now gets extra $150($375 total) through November 30
Sign Up Now: Perchai.app
Also if you like the idea of what I said you can try us free of cost at: perchai.app
Perch now speaks your language. On web and desktop, the app is now available in 7 languages. Simplified and Traditional Chinese, Spanish, Japanese, Thai, Vietnamese and Indonesian.
Dictation, in beta, understands 63+ languages, so you can talk to it in yours. Free on every plan.
What language would you like us to add next?
Native speaker and something reads off? We’re looking for one reviewer per language to make Perch sound natural. Reply or DM and we’ll send you a review sheet.
Check it out at perchterminal.com
Alipay and WeChat Pay now work on Perch.
While we are working on making recurring subs work, Get a one-time pass: 30 days of Pro ($10) or Ultra ($30), paid once. Nothing renews. Buy again anytime to add 30 days.
Settings → Plan & usage → One-time pass
perchai.app
Since all of our users seem to love DeepSeek V4.1 Flash 1017, we will be increasing our limits on our plans.
$10 Pro plan now gets extra $50($125 total) of usage on DeepSeek and the $30 Ultra Plan now gets extra $150($375 total) through November 30
Sign Up Now: Perchai.app
Perch's Agent Browser, Now even better
It opens your site, checks it at any page size from desktop down to phone, and fixes what breaks. Every chat now gets its own tab, so several agents can browse at the same time.
In the Perch desktop app: perchterminal.com
Since a lot of y'all asked for it, we are launching the Perch Desktop App on Linux in Beta.
We have also implemented auto-update across all platforms so for this one last time we will ask you to manually update your desktop apps to V2.5.0
Download Perch Desktop: perchai.app/desktop
We are extending our free model promo till Nov 20 on these models so you can continue testing out our app:
1) DeepSeek v4.1 Flash (1017)
2) GLM 5.3 Flash
3) Qwen 3.8 Max
Try it out perchai.app
Let us know if you have any feedback or would like us to add anything!
The Linux build is in early Beta and is being tested and we would love any feedback or bugs faced during it and will work on fixing it ASAP. You can reach out to us here or on our Discord or email us at [email protected]
Sub-sessions are now in Perch, on every plan, free included.
Some jobs are too big for one chat. Hand them off and each one gets a chat of its own: it runs in the background, you can open it and steer it mid-task, and it reports back when it’s done.
Subagents vs sub-sessions: a subagent is a quick helper that hands back one answer. A sub-session is a full chat you can keep talking to, on its own model if you want.
Try it in the Perch desktop app: ask for two fixes “in separate sessions” and click Start on both.
perchai.app
How it works:
• Perch suggests a session, or you ask for one. Nothing starts until you click Start.
• Each runs in the background as its own chat, under the one that started it.
• Open one and type to steer it.
• When it’s done or needs you, it reports back.
We are extending our free model promo till Nov 20 on these models so you can continue testing out our app:
1) DeepSeek v4.1 Flash (1017)
2) GLM 5.3 Flash
3) Qwen 3.8 Max
Try it out perchai.app
Let us know if you have any feedback or would like us to add anything!
MiniMax M3.1 Flash is now available on Perch AI completely free of cost.
Our Free Lineup now includes:
1)MiniMax M3.1 Flash
2) GLM 5.3 Flash
3) DeepSeek V4.1 Flash
4) Beam 1.0
Try now: perchai.app










