Intel / post

Arcee puts rate limits on the free Trinity Large Preview endpoint

Source: x.com
Date: Published Mar 17, 2026

Trinity Large Preview took off on @OpenRouter faster than we anticipated, crossing just short of 3 trillion tokens served in its first 50 days. We’re incredibly proud of that adoption, but that level of scale forces some real compute tradeoffs. To keep the free endpoint healthy and reliably support the next Trinity releases, we are introducing lenient rate limits starting tomorrow, Wednesday. We’d rather monitor usage carefully than let latency and throughput degrade for everyone. Free access is staying for now, and we are also working to bring on more capacity and a paid endpoint for those who need more consistency.

Mar 17, 2026 View on X

Context

Trinity Large Preview, the lightly post-trained variant from Arcee's initial Trinity Large release, crossed just under 3 trillion served on OpenRouter within its first 50 days, adoption Arcee says outpaced what it had planned for.

To keep the free endpoint from degrading in latency and throughput for everyone using it, Arcee introduced what it calls 'lenient' rate limits rather than removing free access outright, while it works to add more serving capacity and a separate paid endpoint for users who need more consistent performance.

Terms in this piece · Glossary

token — The chunk of text a model reads and writes in — roughly three-quarters of a word — and the unit AI usage is billed in.

Read the source x.com