← Back to Artificial Intelligence
Overview
Arcee AI links its free chat interface to the Trinity open-weight model family. Developers can separately download the weights or use an OpenAI-compatible hosted API.[1][2]
Use cases
- Try conversational assistance, structured JSON output and tool-oriented workflows; choose local deployment or a hosted API for applications.[2]
Features
- Trinity uses a sparse mixture of experts. Nano and Mini target smaller deployments, while the Large variants target hosted agent workloads.[2]
Limitations
- The model page distinguishes model context capacity from the smaller served API context, and its FAQ gives a different Large context figure. Confirm the endpoint before relying on a maximum.[2]
Access and availability
Specifications
Verified 2026-10-04. Versions and access terms can change.
| Nano / Mini parameters | Nano: 6 billion total, 1 billion active; Mini: 26 billion total, 3 billion active.[2] |
|---|---|
| Nano / Mini context | 128,000 tokens on the model page.[2] |
| Large parameters | 400 billion total, 13 billion active; hosted context varies by release and endpoint.[2] |
| Weights license | Arcee announced OpenMDW-1.1 for the Trinity family on 29 May 2026; inspect the selected distribution license.[3] |
Public user feedback
Attributed external experiences. These are separate from verified specifications and vendor claims.
No attributable user reviews have been verified for this entry yet.
Links
Sources and evidence gaps
- Arcee AI official siteofficial · checked 2026-10-04
- Arcee Trinity models and deployment optionsofficial · checked 2026-10-04
- Trinity OpenMDW-1.1 license announcementofficial · checked 2026-10-04 · published 2026-05-29
- Current chat quotas, API prices and identifiable chat-specific user feedback have not been verified.
- The vendor model page contains conflicting Large context figures; no single chat limit is asserted.