Mini toolsOrnith models
Browse all models from this model family.
ID
Model
Company
Type
Primary task
Open source
-
Ornith-1.0-397B-FP8 is DeepReinforce's open-source 397B MoE agentic coding model with FP8 quantization. Built on Qwen 3.5 and RL-trained to jointly optimize scaffolds and solutions, it achieves state-of-the-art performance on Terminal-Bench 2.1, SWE-Bench, and NL2Repo. Supports reasoning, tool use, and 262K context. MIT licensed.NewTextReleased 9d ago
-
Ornith-1.0-35B is a 35B Mixture-of-Experts reasoning model for agentic coding, post-trained on Qwen 3.5 using a self-improving RL framework that jointly learns solution rollouts and the task-specific scaffolds guiding them. Supports native function calling and 262K context. Scores 75.6 on SWE-Bench Verified and 64.2 on Terminal-Bench 2.1. MIT licensed.NewCodingReleased 11d ago
-
Open-source 9B-parameter language model specialized for agentic coding tasks. Post-trained on Qwen 3.5 using a self-improving RL framework that jointly learns to generate solutions and task-specific scaffolds. Achieves state-of-the-art results on SWE-bench Verified (69.4%) and Terminal-Bench 2.1 among comparable models. MIT licensed.NewTextReleased 11d ago
-
Ornith-1.0-397B is a 397B MoE open-source reasoning model for agentic coding, post-trained on Qwen 3.5 MoE via a self-improving RL framework that jointly learns task solutions and the scaffolds guiding them. Achieves 82.4 on SWE-Bench Verified and 77.5 on Terminal-Bench 2.1. MIT licensed with tool-calling and 256K context support.NewMultimodalReleased 11d ago
-
Ornith-1.0-35B-FP8 is an FP8-quantized 35B-parameter mixture-of-experts language model specialized for agentic coding. Built on Qwen3.5 MoE, it is trained with a self-improving RL framework that jointly learns to solve coding tasks and generate the scaffolds guiding those solutions. MIT-licensed and available for self-hosting via vLLM, SGLang, and Transformers.NewTextReleased 12d ago
