Polaris-4B-Preview-F32-GGUF

Polaris is an open-source post-training method that uses reinforcement learning (RL) scaling to refine and enhance models with advanced reasoning abilities. Our research shows that even top-tier models like Qwen3-4B can achieve significant improvements on challenging reasoning tasks when optimized with Polaris. By leveraging open-source data and academic-level resources, Polaris pushes the capabilities of open-recipe reasoning models to unprecedented heights. In benchmark tests, our method even surpasses top commercial systems, including Claude-4-Opus, Grok-3-Beta, and o3-mini-high .

Model Files

File Name	Size	Format	Description
Polaris-4B-Preview.F32.gguf	16.1 GB	F32	Full precision 32-bit floating point
Polaris-4B-Preview.F16.gguf	8.05 GB	F16	Half precision 16-bit floating point
Polaris-4B-Preview.BF16.gguf	8.05 GB	BF16	Brain floating point 16-bit
Polaris-4B-Preview.Q8_0.gguf	4.28 GB	Q8_0	8-bit quantized
Polaris-4B-Preview.Q6_K.gguf	3.31 GB	Q6_K	6-bit quantized
Polaris-4B-Preview.Q5_K_M.gguf	2.89 GB	Q5_K_M	5-bit quantized, medium quality
Polaris-4B-Preview.Q5_K_S.gguf	2.82 GB	Q5_K_S	5-bit quantized, small quality
Polaris-4B-Preview.Q4_K_M.gguf	2.5 GB	Q4_K_M	4-bit quantized, medium quality
Polaris-4B-Preview.Q4_K_S.gguf	2.38 GB	Q4_K_S	4-bit quantized, small quality
Polaris-4B-Preview.Q3_K_L.gguf	2.24 GB	Q3_K_L	3-bit quantized, large quality
Polaris-4B-Preview.Q3_K_M.gguf	2.08 GB	Q3_K_M	3-bit quantized, medium quality
Polaris-4B-Preview.Q3_K_S.gguf	1.89 GB	Q3_K_S	3-bit quantized, small quality
Polaris-4B-Preview.Q2_K.gguf	1.67 GB	Q2_K	2-bit quantized

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):