ZMIME

Aurora2-400M-A20M vs DaisyRose-27B

Compare documented capabilities and access. A dash means a comparable figure is unavailable.

Model specifications
SpecificationAurora2-400M-A20MAuroraAI ResearchDaisyRose-27BRadixArk
TaskBase language modelingAgentic coding
Published capabilitiesEnglish base model pretrained on 100 billion tokens with one shared expert and 32 routed experts per layer.Post-trained from Qwen3.8-27B on 1,200 agent tasks. RadixArk reports 12% pass@1 on Terminal-Bench 4.1.
Model size400M total / approximately 19M active parameters27B parameters
AccessPublic Transformers checkpoint for completion and adaptationPublic BF16, FP8 and NVFP4 checkpoints
SourcesModel weightsPublisher documentation
InputsTextText, Images
Context window—262,144 tokens
WeightsOpen weights · Apache-2.0Open weights · Apache-2.0
ReleasedNov 18, 2026Nov 18, 2026