Research Preview

The Treevik Model Family is currently under active research. No Treevik models should be used for commercial purposes or critical applications requiring high intelligence safely.

Currently, our 2.5B parameter model (Treevik 2) performs 80% better than Treevik 1 and generates usable replies. However, they are still not evaluated for production-level reliance.

FAQ

Treevik 2 (2.5B Instruct) is our most capable model to date.

While we are actively developing robust commercial variants, we are not yet sufficiently confident in the safety and hallucination rates of the current experimental models for production environments.

PlusB Intelligence Lab—including product development, system enhancement, and custom model training—is entirely self-funded. Training trillion-parameter models requires immense capital. We are currently researching novel, efficient architectures that punch above their weight class instead of relying purely on scale.

Extended context windows exponentially increase inference costs. We actively made a trade-off to prioritize reasoning quality over context length within our current operational constraints.

info@plusb.in

Loading...

Type your prompt below and press the button to start the journey.

Treevik Models can generate significant wrong information. Learn more About Treevik models.