basal-omni-2.0-mini released with 0.811 Werdykt score and two-way state reading

·2026.11.15 00:16

Key point

The early release of the small model on Gemma 4 E4B scores 0.811 on the Werdykt benchmark and requires basal engine 2.8.0.

Early release of basal-omni-2.0-mini, a small model based on Gemma 4 E4B, is available with a 0.811 score on the Werdykt benchmark (5,000 hidden decisions in Polish and English).
The model introduces two-way reading, processing state in both directions before answering, which requires basal engine 2.5 or later.
It runs on the public basal engine 2.8.0 and maintains the same API as basal-omni-1.0.
The model outputs a probability distribution over option codes (A-Z, then two-character codes, up to 255 options) without text generation, ensuring answers stay within provided options.

This summary was generated automatically by AI. Check the original for the author's claims and context. Copyright belongs to the original author.

Our guide explains how the AI works. Report summary errors, attribution issues, or removal requests via Contact.