kaitchup/

Olmo-3-7B-Think-fp8-dynamic

This is allenai/Olmo-3-7B-Think quantized with LLM Compressor with FP8 Dynamic (W8A8). The model is compatible with vLLM (tested: v0.11.2). Tested with an RTX 5090.

How the models perform (token efficiency, accuracy per domain, ...) and how to use them: Quantizing Olmo 3: Most Efficient and Accurate Formats

Subscribe to The Kaitchup. This helps me a lot to continue quantizing and evaluating models for free. Or you can "buy me a kofi".