Unverified50% confidenceFactExact time
MNN-Pocket-0.3B is a 300M-parameter dense model.
1
Sources
50%
Confidence
Long-term
Relevance
11/10/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedMNN-Pocket-0.3B supports a context length of up to 32K tokens.70% similarUnverifiedEdge scenarios typically choose small models with 1B to 3B parameters, where a 3B model is about 1.5GB after 4-bit quantization and a 1B model is about 500MB.65% similarUnverified"3B" refers to a model with approximately 3 billion parameters, a scale whose hardware requirements are far lower than those of larger models such as 7B and 13B.65% similarVerifiedAfter applying INT4 quantization, the memory footprint of a 3B parameter model can be reduced to approximately 2 GB.64% similarUnverifiedUnsloth's 3bit quantization is not a standard bitsandbytes configuration; it usually refers to its own asymmetric quantization scheme, which can compress 7B-class model weights to about 3GB.63% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1154758API
curl https://kongchang.com/api/v1/knowledge/claims/1154758MCP
get_claim(id=1154758)