OpenELM 3B

OpenELM 3B needs an estimated 2.6 GB of VRAM at Q4, 4.2 GB at Q8 and 7.5 GB at FP16. All figures are estimates derived from the parameter count, covering weights plus about 15% and 0.5 GB for a short context.

Yes. At FP16, with an estimated 7.5 GB of VRAM, it fits on a 24 GB, 16 GB, 12 GB or 8 GB GPU.

Yes. At Q4 it needs an estimated 2.6 GB, and assuming macOS lets the GPU use about 75% of unified memory, the smallest common Mac memory size that fits is 8 GB.

Yes. OpenELM 3B has open weights, so you can self-host it for full control of your data.

Run OpenELM 3B locally

No. OpenELM 3B only takes text as input and cannot read images, audio, video or PDFs.

The closest alternatives are OLMo 1.7 7B by Ai2, Hermes 2 Pro Llama 3 8B by Nous and Mistral 7B Instruct v0.3 by Mistral AI. They are matched on model kind and release date, with no comparable price.

OLMo 1.7 7BOpenELM 3B vs OLMo 1.7 7BHermes 2 Pro Llama 3 8BOpenELM 3B vs Hermes 2 Pro Llama 3 8BMistral 7B Instruct v0.3OpenELM 3B vs Mistral 7B Instruct v0.3

Open weights. OpenELM 3B is not fully open source, but its weights are available under the apple-amlr license.

The OpenELM 3B weights can be downloaded from the Hugging Face repository apple/OpenELM-3B.

Run OpenELM 3B locally

OpenELM 3B was made by Apple and released on April 23, 2024.

More from Apple

Yes. OpenELM 3B has open weights, so there is no charge from Apple for the model itself.