Model
Olmo3-7B-Instruct
A 7-billion-parameter instruction-tuned version of the open-source Olmo series of large models released by AI2, intended for research and training experiments.
Timeline (last 90 days)
Oct 15
When FSPO is trained on Olmo3-7B-Instruct, peak memory drops from about 70GB to 56GB, saving 14GB.
Unverified50%