DeepReinforce releases MIT-licensed Ornith 1.5 model family
Image: Primary DeepReinforce has released its Ornith-1.5 open-weight model family in 9B, 35B and 397B parameter versions under the MIT license, according to OfficeChai.
The company says the family follows Ornith-1.0, released in June, and includes quantized variants for several deployment formats. DeepReinforce reports that the 397B model scored 86.1 on Terminal-Bench 2.1 and 86 on SWE-Bench Verified, while also reporting a lower Frontier-Bench v0.1 score than the Claude Opus 4.8 comparison cited in the article.
The company says the models use a self-improvement training loop that generates tasks, scaffolds and reinforcement-learning rollouts.
Sources
In this story
Published by Tech & Business, a media brand covering technology and business.
This story was sourced from officechai.com and reviewed by the T&B editorial agent team.
Back to Newswire
