qwen3-4b-code-reasoning-f32-GGUF

Qwen 3 Code Reasoning is a compact 4-billion-parameter model fine-tuned on nvidia/OpenCodeReasoning, designed specifically for coding and logical reasoning tasks. It excels at code generation and logical question answering. To get the best results, it is recommended to use a context window greater than 10,000 tokens, with a temperature of 0.7, top-p of 60, top-k of 0.99, and a minimum p of 0.05. Running the model in "thinking mode" is advised for more coherent and insightful outputs. While powerful, the model may occasionally overthink simple prompts and can run into context overflow with very long inputs.

Model Files

File Name	Quant Type	Size
qwen3-4b-code-reasoning.BF16.gguf	BF16	8.05 GB
qwen3-4b-code-reasoning.F16.gguf	F16	8.05 GB
qwen3-4b-code-reasoning.F32.gguf	F32	16.1 GB
qwen3-4b-code-reasoning.Q2_K.gguf	Q2_K	1.67 GB
qwen3-4b-code-reasoning.Q3_K_L.gguf	Q3_K_L	2.24 GB
qwen3-4b-code-reasoning.Q3_K_M.gguf	Q3_K_M	2.08 GB
qwen3-4b-code-reasoning.Q3_K_S.gguf	Q3_K_S	1.89 GB
qwen3-4b-code-reasoning.Q4_K_M.gguf	Q4_K_M	2.5 GB
qwen3-4b-code-reasoning.Q4_K_S.gguf	Q4_K_S	2.38 GB
qwen3-4b-code-reasoning.Q5_K_M.gguf	Q5_K_M	2.89 GB
qwen3-4b-code-reasoning.Q5_K_S.gguf	Q5_K_S	2.82 GB
qwen3-4b-code-reasoning.Q6_K.gguf	Q6_K	3.31 GB
qwen3-4b-code-reasoning.Q8_0.gguf	Q8_0	4.28 GB

Quants Usage

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

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