Model Detail

CogEvol-4B is a large language model with 4B parameters released by CogEvol. The model is registered under the text-generation pipeline tag on Hugging Face, distributed under the permissive apache-2.0 license.

CogEvol-4B ships with 4B parameters. Total weight footprint is approximately 4.5 GB, which is the relevant figure when planning local-inference VRAM. The apache-2.0 license is permissive, allowing commercial deployment and derivative work without per-seat fees, though attribution requirements still apply.

CogEvol-4B is best fit for general-purpose chat and instruction-following workloads. Treat this as a starting matrix rather than a benchmark verdict — the right deployment usually depends on the specific evaluation suite that mirrors your workload.