OLMoE 1B-7B January 2025 Instruct
This is the instruction-tuned January 2025 OLMoE mixture-of-experts checkpoint.
What the release provides
Its post-training sequence includes supervised fine-tuning, preference optimization and verifiable-reward training. Ai2 also publishes the preceding stages and a reward model. The instruction weights use Apache 2.0.OLMoE 1B-7B January 2025 Instruct official model card
How to use it well
Practical guidance. Account for the full stored model when sizing memory. Active compute and total weight storage are different costs, so measure both on the target device.
Specifications & access
| Detail | Value and conditions |
|---|---|
| Access | Downloadable |
| Model identifiers | allenai/OLMoE-1B-7B-0125-Instruct |