Trinity Large Thinking
Trinity Large Thinking is Arcee AI’s reasoning-focused post-training of Trinity Large Base. The sparse mixture-of-experts model has 398B total parameters and approximately 13B active per token, with documented context extension to 512K. It generates explicit thinking traces and is trained for long-running agents, multi-turn tool calls, and coherent instruction following. Thinking tokens must remain in conversation history for supported agent behavior. API access and downloadable weights are available. The current model card specifies OpenMDW-1.1, despite Apache 2.0 wording in the original announcement; this record follows the current weight license.
2026-04-01
398B total, approximately 13B active
Sparse Mixture-of-Experts Transformer
OpenMDW-1.1
Specifications
- Parameters
- 398B total, approximately 13B active
- Architecture
- Sparse Mixture-of-Experts Transformer
- License
- OpenMDW-1.1
- Context Window
- 524,288 tokens
- Type
- text
- Modalities
- text
Benchmark Scores
Advanced Specifications
- Model Family
- Trinity
- Finetuned From
- Trinity Large Base
- API Access
- Available
- Chat Interface
- Available
- Variants
- Trinity-Large-ThinkingW4A16
Capabilities & Limitations
- Capabilities
- reasoningagentic codingtool callinglong context
- Known Limitations
- Thinking traces must be preserved across turnsLarge total weights require substantial memoryCurrent weight license differs from original blog wording
- Notable Use Cases
- long-running local agentsprivate workflow automationreasoning research
- Function Calling Support
- Yes
- Tool Use Support
- Yes
