Model Select
Query-aware routing. Telluvian/Gallery-1 picks the best model for each prompt. How it works
Route,Verify,Optimise.
We make AI cheaper, easier, and more reliable.
Model-agnostic Minimal added latency OpenAI compatible API
Backed by investors and Angels from:
What Telluvian does:
Verification
A probe reads the model's hidden states as it generates, detecting errors in real-time. How token-level detection works
95%+accuracy detecting GPT-5.4 hallucinations
how many ‘a’s are there in: antidisestablishmentarianism
There are 3 “a”s in antidisestablishmentarianism.
Optimisation
Same answer quality, optimised cost. Finding the cost–accuracy optimum
Interested? We're hiring.
Open roles across research, engineering and go-to-market.
Curious about mechanistic interpretability? Try our live J-space demo →