NanoGPTDaisyRose 27B

NanoGPT

DaisyRose 27B

DaisyRose 27B is RadixArk’s multimodal reasoning model for coding, tool use and long-horizon agent tasks. Built from Qwen3.8-27B using reinforcement learning without supervised fine-tuning or distillation, it accepts text and images and returns text. It supports function calling, JSON-schema output and a 262,144-token context window, with up to 65,536 output tokens. Thinking is on by default; choose low, medium or high effort, or turn it off. The weights are released under Apache 2.0.

radixark/daisyrose-27b
Context
262.1K
Max output
65.5K
Input $/M
$0.15
Output $/M
$1.88
Cache read $/M
$0.037
Cache write $/M
—

Against the catalogue

Context window262.1K

Larger than 44% of the 10,022 models listed

Input price$0.15

Cheaper than 74% of priced models

Capabilities

Reasoningsupported
Tool callingsupported
Structured outputsupported
Attachmentssupported
Temperaturenot supported
Open weightssupported
Interleaved thinkingnot supported

Modalities

Input
TextImage
Output
Text

Reasoning controls

effortnone, low, medium, high

Pricing detail

RateBase $/M
Input$0.15
Output$1.88
Cache read$0.037