Radixark API: models, pricing, and specs
Compare every Radixark model available through Felo API — prices, context windows, and capabilities in one table. Call any of them with the same API key.
- Models
- 1
- chat ×1
- Input per 1M
- $0.2
- Output per 1M
- $2.5
- Largest context
- 262.144K
RadixArk: DaisyRose27B is a multimodal reasoning model tuned for coding, tool use, and long-horizon agent work, with a262,144-token context.
RadixArk: DaisyRose27B is a multimodal reasoning model based on Qwen3.8-27B. It accepts text and images, and has a262,144 tokens (256K) context window. Its training used reinforcement learning only, without supervised fine-tuning or distillation, while preserving the base model’s architecture, tokenizer, and chat template.
The model is designed for coding, tool use, and long-horizon agent tasks. It is suited to software engineering, machine learning, and science problem-solving, as well as terminal-based agent workflows, code generation and editing, instruction following, and reasoning. Its large context can support long-document processing and analysis of large codebases.
On every listed benchmark, DaisyRose27B scored higher than the base model. Its Terminal-Bench4.1 pass@1 score was12.0%, compared with5.6% for the base model. These results make it relevant for teams evaluating agentic coding and complex problem-solving workflows; they do not establish performance on tasks beyond those reported benchmarks.
Specifications
- Input per 1M
- $0.2
- Output per 1M
- $2.5
- Context
- 262.144K
- Max output
- 235.929K
- Protocols
- messages