ModelOrnith 1.5 / Ornith
Ornith 1.5 35B-A3B
A large language model using a MoE (mixture-of-experts) architecture with 35B total parameters and about 3B parameters activated per token. Combined with DFlash speculative decoding, it enables high-speed local inference and targets low first-token latency scenarios.
Core Facts
Timeline (last 90 days)
Oct 12
在Mac Studio M4 Max（64GB）上，Ornith 1.5 35B-A3B的推理速度全面优于Qwen3.8-27B，但两者代码审查能力基本持平。
Unverified50%
Aug 30
Ornith 1.5 35B-A3B uses a MoE architecture with a total parameter count of approximately 35B, activating only approximately 3B parameters per inference.
Verified65%