metadata
license: apple-ascl
OpenELM 3B Instruct GGUFs
After a long wait, Llama.cpp (b3324) finally supports OpenELM, and that means OpenELM GGUFs! iMatrix quants are coming soon!
Download the GGUFs below!
| Filename | Quant type | Description |
|---|---|---|
| OpenELM-3B-Instruct-F32.gguf | F32 | Full precision, 32-bit floating point. Largest file size, baseline quality. |
| OpenELM-3B-Instruct-F16.gguf | F16 | Full precision, 16-bit floating point. Very large file size, highest quality. |
| OpenELM-3B-Instruct-Q8_0.gguf | Q8_0 | Extremely high quality, generally unneeded but max available quant. |
| OpenELM-3B-Instruct-Q6_K.gguf | Q6_K | Very high quality, near perfect, recommended. |
| OpenELM-3B-Instruct-Q5_K_M.gguf | Q5_K_M | High quality, recommended. |
| OpenELM-3B-Instruct-Q5_K_S.gguf | Q5_K_S | High quality, recommended. |
| OpenELM-3B-Instruct-Q5_1.gguf | Q5_1 | High quality, improved 5-bit quantization. |
| OpenELM-3B-Instruct-Q5_0.gguf | Q5_0 | High quality, alternative 5-bit quantization. |
| OpenELM-3B-Instruct-Q4_K_M.gguf | Q4_K_M | Good quality, recommended. |
| OpenELM-3B-Instruct-Q4_K_S.gguf | Q4_K_S | Slightly lower quality with more space savings, recommended. |
| OpenELM-3B-Instruct-Q4_1.gguf | Q4_1 | Good quality, improved 4-bit quantization. |
| OpenELM-3B-Instruct-Q4_0.gguf | Q4_0 | Good quality, older 4-bit quantization. |
| OpenELM-3B-Instruct-IQ4_XS.gguf | IQ4_XS | Decent quality, smaller than Q4_K_S with similar performance, recommended. |
| OpenELM-3B-Instruct-Q3_K_L.gguf | Q3_K_L | Lower quality but usable, good for low RAM availability. |
| OpenELM-3B-Instruct-Q3_K_M.gguf | Q3_K_M | Even lower quality. |
| OpenELM-3B-Instruct-IQ3_M.gguf | IQ3_M | Medium-low quality, new method with decent performance comparable to Q3_K_M. |
| OpenELM-3B-Instruct-IQ3_XS.gguf | IQ3_XS | Lower quality, new method with decent performance, slightly better than Q3_K_S. |
| OpenELM-3B-Instruct-Q3_K_S.gguf | Q3_K_S | Low quality, not recommended. |
| OpenELM-3B-Instruct-IQ3_XXS.gguf | IQ3_XXS | Lower quality, new method with decent performance, comparable to Q3 quants. |
| OpenELM-3B-Instruct-Q2_K.gguf | Q2_K | Very low quality but surprisingly usable. |
| OpenELM-3B-Instruct-IQ2_M.gguf | IQ2_M | Very low quality, uses SOTA techniques to also be surprisingly usable. |
| OpenELM-3B-Instruct-IQ2_S.gguf | IQ2_S | Very low quality, uses SOTA techniques to be usable. |
| OpenELM-3B-Instruct-IQ2_XS.gguf | IQ2_XS | Very low quality, uses SOTA techniques to be usable. |
| OpenELM-3B-Instruct-IQ2_XXS.gguf | IQ2_XXS | Extremely low quality, smallest 2-bit quantization. |
| OpenELM-3B-Instruct-IQ1_M.gguf | IQ1_M | Extremely low quality, 1-bit quantization. |
| OpenELM-3B-Instruct-IQ1_S.gguf | IQ1_S | Extremely low quality, smallest possible 1-bit quantization. |