Dhīr 5B MoE vs DaisyRose-27B vs EKA-1
Compare documented capabilities and access. A dash means a comparable figure is unavailable.
| Specification | |||
|---|---|---|---|
| Task | Multilingual assistant | Agentic coding | Multilingual reasoning and coding |
| Published capabilities | Mixture-of-experts assistant supporting English and ten Indian languages for offline deployment. | Post-trained from Qwen3.8-27B on 1,200 agent tasks. RadixArk reports 12% pass@1 on Terminal-Bench 4.1. | Trained on about 7.5 trillion tokens across English and all 22 scheduled Indian languages. |
| Model size | 5.09B total / 1.79B active | 27B parameters | 26.3B parameters |
| Access | Open weights for on-premises or offline use | Public BF16, FP8 and NVFP4 checkpoints | Released through the IndiaAI Mission |
| Sources | Publisher documentation · Additional documentation | Publisher documentation | Publisher documentation |
| Inputs | Text | Text, Images | Text |
| Context window | 16,384 tokens | 262,144 tokens | — |
| Input price per million tokens | — | — | — |
| Output price per million tokens | — | — | — |
| Weights | Open weights | Open weights · Apache-2.0 | Weights unconfirmed |
| Released | Nov 3, 2026 | Nov 18, 2026 | Nov 3, 2026 |