DeepReinforce.AI Ornith 1.5

Ornith 1.5 397B is the flagship 397B parameter Mixture-of-Experts model, designed for advanced coding and agentic workloads. Its end-to-end self-improvement approach jointly optimizes task generation, scaffold construction, and solution rollouts through reinforcement learning. The Qwen3.5-based architecture features 60 layers with hybrid linear and full attention, 32 attention heads, 2 KV heads, and 512 experts with 10 activated per token. The model supports a 262,144 token context window.

Ornith 1.5 35B A3B is the mid-size Mixture-of-Experts member of the Ornith-1.5 family, designed for coding and agentic workloads with only ~3B parameters activated per token. Its end-to-end self-improvement approach jointly optimizes task generation, scaffold construction, and solution rollouts through reinforcement learning. The Qwen3.5-based architecture features 40 layers with hybrid linear and full attention, 16 attention heads, 2 KV heads, and 256 experts with 8 activated per token. The model supports a 262,144 token context window.

Ornith 1.5 9B is the lightweight member of the Ornith-1.5 family, designed for efficient single-GPU deployment and edge use through its quantized mobile variant. It is a 9B dense model built around an end-to-end self-improvement approach that jointly optimizes task generation, scaffold construction, and solution rollouts through reinforcement learning. Its Qwen3.5-based architecture uses 32 layers with hybrid linear and full attention, 16 attention heads, and a 4-head KV configuration. The model supports a 262,144 token context window, making it suitable for efficient deployment across demanding long-context workloads.