angrygiraffe/claude-opus-4.6-4.7-reasoning-8.7k
Viewer • Updated • 38.5k • 1.68k • 440
How to use mondk/Safetensors.Claude-Qwen3.5-4B-Reasoning with Unsloth Studio:
curl -fsSL https://unsloth.ai/install.sh | sh # Run unsloth studio unsloth studio -H 0.0.0.0 -p 8888 # Then open http://localhost:8888 in your browser # Search for mondk/Safetensors.Claude-Qwen3.5-4B-Reasoning to start chatting
irm https://unsloth.ai/install.ps1 | iex # Run unsloth studio unsloth studio -H 0.0.0.0 -p 8888 # Then open http://localhost:8888 in your browser # Search for mondk/Safetensors.Claude-Qwen3.5-4B-Reasoning to start chatting
# No setup required # Open https://huggingface.co/spaces/unsloth/studio in your browser # Search for mondk/Safetensors.Claude-Qwen3.5-4B-Reasoning to start chatting
pip install unsloth
from unsloth import FastModel
model, tokenizer = FastModel.from_pretrained(
model_name="mondk/Safetensors.Claude-Qwen3.5-4B-Reasoning",
max_seq_length=2048,
)datasets:
It will no longer identify itself as claude.
Loss: 1.4
Better coding: +23%
gguf: mondk/GGUF.Claude-Qwen3.5-4B-Reasoning
ty
The questions used in the Package/question_set.txt (generated by Claude—it's in Vietnamese, bruh).
rated by gemini
temp = 0.0, system prompt:
You are a precise coding assistant. Solve the problem exactly as specified.
Output ONLY the final code in a single code block, with no explanation,
no preamble, and no comments unless explicitly requested. Follow the
function signature exactly if one is given.
..................
Results:
| # | Category | Base pass? | Fine-tuned pass? |
|---|---|---|---|
| E1-E5 | Easy | 5/5 | 5/5 |
| M1-M8 | Medium | 6/8 | 8/8 |
| H1-H7 | Hard | 4/7 | 5/7 |
| Total | 15/20 | 18/20 |