Crux-Qwen3_OpenThinking-4B-GGUF

Crux-Qwen3_OpenThinking-4B is fine-tuned on the Qwen3-4B architecture, optimized for advanced open thinking, mathematical reasoning, and logical problem solving. This model is trained on the traces of sk1.1, which include 1,000 entries from the Gemini thinking trajectory, combined with fine-tuning on 100k tokens of open math reasoning data. This makes it highly effective for nuanced reasoning, educational tasks, and complex problem-solving requiring clear thought processes.

Model Files

File Name	Size	Quantization	Format	Description
Crux-Qwen3_OpenThinking-4B-GGUF.F16.gguf	8.05 GB	FP16	GGUF	Float16 precision version
Crux-Qwen3_OpenThinking-4B-GGUF.Q4_K_M.gguf	2.5 GB	Q4_K_M	GGUF	4-bit quantized (K M variant)
Crux-Qwen3_OpenThinking-4B-GGUF.Q5_K_M.gguf	2.89 GB	Q5_K_M	GGUF	5-bit quantized (K M variant)
Crux-Qwen3_OpenThinking-4B-GGUF.Q8_0.gguf	4.28 GB	Q8_0	GGUF	8-bit quantized
.gitattributes	1.88 kB	—	—	Git LFS tracking file
config.json	31 B	—	—	Configuration file
README.md	31 B	—	—	Model documentation

Quants Usage

(sorted by size, not necessarily quality. IQ-quants are often preferable over similar sized non-IQ quants)

Link	Type	Size/GB	Notes
GGUF	Q2_K	0.4
GGUF	Q3_K_S	0.5
GGUF	Q3_K_M	0.5	lower quality
GGUF	Q3_K_L	0.5
GGUF	IQ4_XS	0.6
GGUF	Q4_K_S	0.6	fast, recommended
GGUF	Q4_K_M	0.6	fast, recommended
GGUF	Q5_K_S	0.6
GGUF	Q5_K_M	0.7
GGUF	Q6_K	0.7	very good quality
GGUF	Q8_0	0.9	fast, best quality
GGUF	f16	1.6	16 bpw, overkill

Here is a handy graph by ikawrakow comparing some lower-quality quant types (lower is better):