DaisyRose-27B-NVFP4-GGUF

DaisyRose-27B is RadixArk's debut open-weight multimodal reasoning model (27.8B total parameters), built upon Qwen3.8-27B using pure reinforcement learning via the open-source Miles framework without any supervised fine-tuning or distillation. Trained over 18 days on 144 H200 GPUs across roughly 1,200 curated long-horizon agent tasksβ€”featuring multi-step sandbox sessions of up to 1M tokens and over 100 tool callsβ€”it more than doubles its base model's Terminal Bench 4.1 pass@1 score (rising from 5.6% to 12.0%) while boosting performance across SWE Bench Pro v2, DeepSWE, IFBench, and GPQA Diamond. Because it preserves Qwen3.8-27B's hybrid linear/full-attention architecture, vision encoder, tokenizer, and 256K-token context window, it serves as a direct drop-in replacement in existing stacks (such as SGLang with DSpark speculative decoding) and is available under an Apache 2.0 license in BF16, FP8, and NVFP4 formats.

MTP β€” Speculative Decoding: Yes

Multi-Token Prediction (MTP) GGUF is a specialized GGUF model file format extension that integrates speculative decoding directly into the model weights to significantly accelerate local inference. Unlike traditional speculative decoding which requires a separate, smaller "draft" model, MTP GGUF files include additional output heads within the main model architecture that predict multiple future tokens in a single forward pass.

Model Files

File Name Quant Type File Size File Link Description
DaisyRose-27B-NVFP4.gguf NVFP4 21 GB Link NVFP4 quantization. Highly optimized for NVIDIA Blackwell hardware accelerators (e.g., RTX Pro 6000, RTX 50-series, B200, B300), offering native hardware-accelerated FP4 Tensor Core speeds and a strong balance of size and performance for this 27B parameter model.
DaisyRose-27B.mmproj.gguf mmproj 931 MB Link Multimodal projection file. Used for vision/language capabilities.

NVFP4 Test Run

Tested DaisyRose-27B-NVFP4.gguf on an NVIDIA RTX PRO 6000 Blackwell Server Edition GPU with 96 GB of VRAM, using llama-cli with full GPU offloading and MTP speculative decoding.

Performance Metrics

Metric Value
Average Prompt Processing Speed 144.66 tokens/s
Average Generation Speed 114.34 tokens/s
Total Wall Time 56.4 seconds
Context Length 16,384 tokens
KV Cache (K) q8_0
GPU NVIDIA RTX PRO 6000 Blackwell Server Edition
GPU VRAM Capacity 96 GB
Peak GPU VRAM Usage 19.69 GiB
Peak GPU Utilization 96%
GPU Layers Offloaded 99 (-ngl 99)
Random Seed 42
Speculative Decoding MTP (draft_n_max=2)
Test Runs 5

Test Shots

Shot Test
Shot 1 C: Insert Node at End of Singly Linked List
Shot 2 Jogger and Train Passing Time
Shot 3 Two Pipes Filling a Tank
Shot 4 Students in Examination Rooms
Shot 5 Father and Son Ages

Run Inference

./build/bin/llama-cli \
  -m /path/to/models/DaisyRose-27B-NVFP4.gguf \
  --spec-type draft-mtp \
  --spec-draft-n-max 2 \
  -ngl 99 \
  -c 16384 \
  --cache-type-k q8_0 \
  --seed 42 \
  -n 8192 \
  -st \
  -p "Your prompt here"

Performance results reflect this specific hardware and inference configuration. Actual throughput may vary depending on the prompt, model, software build, and runtime settings.

Snapshots (5)

download (1) download (2) download (3) download (4) download (5) download

llama.cpp

LLM inference in C/C++ β€” https://github.com/ggml-org/llama.cpp

Quantized using llama.cpp by prithivMLmods.

Downloads last month
61
GGUF
Model size
27B params
Architecture
qwen35
Hardware compatibility
Log In to add your hardware

4-bit

Inference Providers NEW
This model isn't deployed by any Inference Provider. πŸ™‹ Ask for provider support

Model tree for prithivMLmods/DaisyRose-27B-NVFP4-GGUF

Base model

Qwen/Qwen3.8-27B
Quantized
(36)
this model

Collections including prithivMLmods/DaisyRose-27B-NVFP4-GGUF