Modelpublic

hudsongouge/minicpm5-1B-GLM-5.2-Agentic-v3

sourceHugging Faceotherupdated 3mo agoView on Hugging Face

0likes451downloads

Model Card

MiniCPM5-1B-Agentic-v3

Created by GLM-5.2. Model 3 of 8 in the agentic post-training series.

Variant: RFT v1 (best individual)

Evaluation

Metric	Score
Real-World Tasks	29.9% (3.59/12)
Unique Tasks Solved	6/12
Consistent (5/5)	rwfixscraper, rwhttpserver

GGUFs available: f16, q80, q5km, q4km, q3km, q2k

Quantization Recommendations

This is a 1B model — heavier quantization degrades output quality significantly.

Quant	Quality	Size	Recommendation
f16	Full	~2.1GB	Best quality
q8_0	Excellent	~1.1GB	Recommended — near-identical to f16
q5km	Good	~0.8GB	Reasoning OK, response may degrade on longer outputs
q4km	Fair	~0.7GB	Reasoning OK, response degrades into repetition
q3km	Poor	~0.6GB	Not recommended
q2_k	Poor	~0.5GB	Not recommended

For production use, prefer q8_0 or f16. The model uses reasoning tokens; lower quantizations break the transition from reasoning to response.

Chat Template

The GGUF chat template defaults to enable_thinking=true, so the model will always produce reasoning followed by response. If your inference engine supports enable_thinking=false, you can skip reasoning for faster responses.