© 2026 — Platform
: The "+Repack" part could refer to a process or feature that repackages the model in some way. This might involve rearranging or optimizing the model's structure for better performance, compatibility, or efficiency on specific hardware or software platforms.
| Tag in Filename | Bits | File Size (7B) | RAM Usage | Quality | Best For | | :--- | :--- | :--- | :--- | :--- | :--- | | | 2-bit | 1.8GB | 2.5GB | Poor | Embedded systems | | q4_0 | 4-bit | 3.8GB | 4.5GB | Good | Old laptops (4GB RAM) | | q4_K_M | 4-bit (K-quant) | 4.1GB | 5GB | Very Good | Best balance | | q5_K_M | 5-bit | 4.7GB | 6GB | Excellent | Desktop CPUs | | q8_0 | 8-bit | 7.3GB | 9GB | Near-lossless | High-end workstations |
In the LLM world, .bin typically refers to a raw binary file containing the model weights. Unlike safetensors (which are also binary but have metadata protection), a .bin might be a direct memory dump of the model state.
