What it is
OLMoE is a fully open-source mixture-of-expert language model designed for efficiency, featuring 1 billion active and 7 billion total parameters. It is optimized for performance on common edge devices and comes with open data, code, and training resources.
Gabriel’s notes
Ai2 and Contextual AI released OLMoE, a first-of-its-kind fully open-source mixture-of-expert (MoE) language model with 1 billion active and 7 billion total parameters that that beats comparable LLMs and can be run easily on common edge devices. OLMoE is pre-trained from scratch and released with open data, code, logs, and intermediate training checkpoints.
Good fit if you want to:
- generate, edit, or enhance creative assets (images, design, branding).
- build, test, or ship software faster (APIs, dev tooling, code assistance).
Pricing snapshot (auto-enriched): Free tier available with $25 in free credits; usage-based pricing for on-demand plans; custom pricing for enterprise plans; pricing includes per-page and per-token rates for various components with no explicit hidden limits mentioned.
Work-use / compliance snapshot (auto-enriched): OLMoE is a fully open-source language model primarily designed for research and development, without explicit built-in workplace compliance features such as data handling policies, training usage restrictions, data retention, SSO availability, or certifications like SOC2, HIPAA, or GDPR.
Alternatives (auto-enriched): Alternative: JetMoE | Comparison: JetMoE is a larger MoE model, but OLMoE achieves better performance with fewer active parameters and less compute cost.
Note: pricing and policy details can change—verify on the official site before making decisions.