Ornith-1.5: AI That Trains Itself by Proposing Its Own Tasks
Ornith-1.5: From Self-Scaffolding to Self-Improvement
Ornith-1.5, a new family of open-source models from Ornith AI, introduces a self-improvement loop where the model generates its own tasks, scaffolds, and solutions for reinforcement learning. The 397B MoE model matches Claude Opus 4.8 on Terminal-Bench 2.1 and DeepSWE, while the 9B edge model outperforms much larger models like Gemma 4-31B. This approach moves beyond static training data, enabling continuous capability gains in coding, reasoning, and agentic tasks.
Instead of relying on a static training distribution or hand-engineered agent design, Ornith-1.5 continually expands its own curriculum and adapts its problem-solving strategies, driving sustained capability gains across reasoning, coding, and agentic tasks.
- montroser
Hoping this is real. It's too bad to see the signals from Qwen that they will not be releasing a 35B-A3B for the 3.8 lineup. The MoE architecture makes a huge difference for being able to run these local models on reasonable consumer hardware.
- lsb
The page has comparisons with Qwen 3.6 27b and I’d love to see comparisons with Qwen 3.8 27b, the newer one is much more capable!
- hxii
Interestingly, in my own benchmark and testing (in the hopes of finding a good-enough local model to run a personal assistant agent), Ornith-1.0-9B was worse than Qwen3.5-9B which according to their scores should've been reversed.
I will definitely pass Ornith-1.5-9B through the gauntlet as well!
- prometheus1992
Can't wait to try this. Ornith1 (9B) was a really nice model. I have been running it locally using - https://github.com/deepanwadhwa/samosa-chat
- rbanffy
It’s time for me to upgrade the main server in my home lab and I’m thinking about which machine should I have.
What kind of hardware you’d need to run the 397B one at an acceptable speed?