Open-weights family of agentic coding models (9B to 397B) that learns to generate its own task-specific scaffolds.
Updated EU status checked Last shown
Sources: the vendor's own pages (linked above) and creator videos at the linked timestamps. EU status is information, not legal advice; see how we check.
Ornith-1.0 is a family of open-source models for agentic coding, built on pretrained Gemma 4 and Qwen 3.5 bases. It comes in four variants: 9B Dense, 31B Dense, 35B MoE and 397B MoE, covering edge deployment up to frontier scale.
Its main training idea is a self-improving RL framework. In each step the model first proposes a refined scaffold (harness) for a task, then generates a solution rollout conditioned on that scaffold. Reward from the rollout is propagated to both stages, so the model learns to author the orchestration as well as the answer. To limit reward hacking, the environment, tool surface and test isolation are fixed outside the model's control. A deterministic monitor assigns zero reward to violations, and a frozen LLM judge acts as a veto. Training uses asynchronous pipeline RL with a staleness-weighted token-level GRPO loss.
The vendor reports that the 397B model scores 77.5 on Terminal-Bench 2.1 and 82.4 on SWE-Bench Verified. It compares these figures with Claude Opus 4.7, MiniMax M3 and DeepSeek-V4-Pro. The 9B model is reported at 43.1 on Terminal-Bench 2.1 and 69.4 on SWE-Bench Verified. These are self-reported benchmark results. Per YouTube coverage, weights and GGUF quantizations are on Hugging Face, but no repository link or license was in the provided material.
Watch the moment: the exact second top creators show it. Videos load only when you press play, via youtube-nocookie.com.
Open-source MoE family (9B to 397B) for agentic coding that learns to generate its own harness/scaffold. Reported benchmarks are on par with larger open models like DeepSeek V4 and GLM; weights and GGUF quants are on Hugging Face.
Closest neighbours by meaning, not just by category.
Model
Open model family (9B dense, 35B MoE, 397B MoE) trained with a self-improvement loop for reasoning, coding and agentic tasks.
Model · Meituan
Meituan's open-weights 1.6T-parameter multimodal MoE model (48B active) for agentic coding and visual tasks.
Model · Accio Lab
Open-weights 35B MoE (3B active) agentic model from Accio Lab, post-trained from Qwen3.6-35B-A3B.
Model · Meituan
Meituan's open-weights 1.6T-parameter mixture-of-experts LLM (48B active), aimed at agentic coding and reasoning tasks.
Model
35B Mixture-of-Experts open-weights model from InternScience for long-horizon agentic search, engineering and tool use.
Model · IQuest Lab
320B-parameter MoE model (15B active) from IQuest for agentic coding and tool use, with 512K context.