Qwen3-8B-CK-Pro-f32-GGUF

The CognitiveKernel/Qwen3-8B-CK-Pro model is a fine-tuned variant of the Qwen3-8B base language model, trained using self-collected trajectories from queries as detailed in the Cognitive Kernel-Pro research. It is designed as a deep research agent and foundation model, achieving strong performance with Pass@1/3 scores of 32.7%/38.2% on the full GAIA dev set and 40.3%/49.3% on the text-only subset. This model builds upon the strengths of Qwen3-8B, which supports advanced reasoning, instruction-following, and multilingual capabilities, specifically optimized for research agent tasks through the Cognitive Kernel-Pro framework. It is not currently deployed by any inference provider on Hugging Face. The model leverages the underlying Qwen3-8B base and its finetuned versions to deliver enhanced agent capabilities for complex question-answering and information synthesis scenarios.

Execute using Ollama

run ->

ollama run hf.co/prithivMLmods/Qwen3-8B-CK-Pro-f32-GGUF:Q2_K

Model Files

File Name	Quant Type	File Size
Qwen3-8B-CK-Pro.BF16.gguf	BF16	16.4 GB
Qwen3-8B-CK-Pro.F16.gguf	F16	16.4 GB
Qwen3-8B-CK-Pro.F32.gguf	F32	32.8 GB
Qwen3-8B-CK-Pro.Q2_K.gguf	Q2_K	3.28 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):