Agent builders who want an open reasoning model they can self-host; Apache 2.0 weights and cheap OpenRouter rates.
Kenning's take
What it does
Trinity-Large-Thinking is a reasoning model from US company Arcee AI, aimed at agents that run for a long time and call many tools. It follows the earlier Trinity-Large-Preview and thinks before answering, which Arcee says improves tool use and instruction following. It is available through Arcee's API and OpenRouter, and the weights are on Hugging Face under the Apache 2.0 licence.
- Open weights under Apache 2.0
- Reasoning before each response
- Multi-turn tool calling for agents
- 262K token context on OpenRouter
- Available via Arcee API and OpenRouter
Who it suits. Developers building coding and task agents who want an open model they can host themselves.
Key facts
- Category
- AI models
- Made by
- Arcee AI
- Based in
- United States
- Released
- 1 April 2026
- How to get it
- Arcee API, OpenRouter, and open weights on Hugging Face
- Works on
- API, open-weights
- In the UK
- Yes
- Free plan
- open weights under Apache 2.0
- API price
- OpenRouter: $0.25 per million input tokens, $0.80 per million output; Arcee's launch post quoted $0.90 per million output on its own API
FAQ
When was Trinity-Large-Thinking released?
Arcee AI released it on 1 April 2026.
Is Trinity-Large-Thinking open source?
The weights are on Hugging Face under the Apache 2.0 licence.
How much does Trinity-Large-Thinking cost?
On OpenRouter it costs $0.25 per million input tokens and $0.80 per million output tokens.
Facts from the maker's own pages and the sources listed. Checked 2026-10-11.
Similar tools
ACE-Step 1.5
Open-source music generation model that writes full songs with lyrics and runs on ordinary consumer GPUs.
AlphaGenome Atlas
Google DeepMind database of AI-predicted effects for every possible single-letter change in human DNA.
Amazon Nova 2
Amazon's second family of Nova AI models, covering reasoning, speech conversation and combined text-and-image generation.
Cisco Antares
Cisco's small open-weight models that find where known security vulnerabilities sit in a codebase, running locally.