SemIf update today: merged Apple Silicon/MPS support (GitHub: dp-IED), a 27B EXL3 bridge hitting 95.8% balanced accuracy (@jkyamog) , and proper temperature calibration (GitHub: samarthpatel24) 📊
Check it out at openjev.com, GitHub.com/theoleecj/semif
On the Jev waitlist?
You can use not Jev (but close enough) right here: github.com/TheoLeeCJ/openjev
Openjev.com browser demo
No waitlist, no funding, no hype, just Open Source Openjev - fast on-device (you can even run it in Chrome!) inference for type-safe decisions with probabilities!
github.com/theoleecj/openjev
openjev.com
Built on Qwen 3 4B / 0.6B / 0.8B.
* no assocation with a Jev model
Theodore retweeted
I think I could achieve similar results with @GoogleDeepMind Gemma3 270M.
After co-inventing ChatGPT, I kept asking myself: why have superhuman chat models not led to AGI?
I’ve spent the last 2 years in stealth building a new way to train models (RLCD), and a new type of frontier AI model that we are releasing today: Jev
• 20-200x faster
• 40-400x cheaper (w/ output tokens free)
• Frontier composable intelligence optimized for decisions
AFAICT the shortest path to AI-based economic revolution
6 hours with Codex and Qwen 3.5 4B (or 0.8B) and you can get something eerily similar - and maybe open source??? github.com/TheoLeeCJ/openjev 👀
Not 2 years in stealth!
It's Not Jev, it's something you can use today - in your browser, even!
After co-inventing ChatGPT, I kept asking myself: why have superhuman chat models not led to AGI?
I’ve spent the last 2 years in stealth building a new way to train models (RLCD), and a new type of frontier AI model that we are releasing today: Jev
• 20-200x faster
• 40-400x cheaper (w/ output tokens free)
• Frontier composable intelligence optimized for decisions
AFAICT the shortest path to AI-based economic revolution
Its accuracy with Qwen 3.5 4B almost matches another currently-closed-source, waitlisted (at time of writing) model on some public evaluations like evals.typesafe.ai/ and typesafe-parallel-judgment-l…, obtaining 0.845 against the closed model's 0.883.
