Model ID
arcee-ai/Trinity-Large-Thinking
Capabilities
Trinity Large Thinking
Trinity Large Thinking is a reasoning-optimized sparse Mixture-of-Experts language model with 398B total parameters and approximately 13B active parameters per token. It features 60 layers, 48 attention heads, and 256 experts, activating 4 experts per token alongside 1 shared expert. The model combines sliding-window and full attention with a 262K-token context window and is optimized for agentic reasoning, multi-step planning, tool use, and long-context workflows, generating explicit reasoning enclosed in dedicated reasoning tags before producing its final response.
About this model
Trinity Large Thinking is a reasoning-optimized sparse Mixture-of-Experts language model with 398B total parameters and approximately 13B active parameters per token. It features 60 layers, 48 attention heads, and 256 experts, activating 4 experts per token alongside 1 shared expert. The model combines sliding-window and full attention with a 262K-token context window and is optimized for agentic reasoning, multi-step planning, tool use, and long-context workflows, generating explicit reasoning enclosed in dedicated reasoning tags before producing its final response.
Model specifications and capabilities are published by the model author and reproduced here from HF Model Card (arcee-ai/Trinity-Large-Thinking). Released under OpenMDW-1.1. Further reading: Paper · Blog
Deploy Private Inference Endpoint
Choose your hardware and inference engine to get deployment commands and performance benchmarks tailored to your infrastructure
Hardware
B200
1,440 GB
Engine
vLLM
SGLang
Deployment code
Access Restricted
Enter your email to get access to this content
Performance under load
Tokens/s (total)
Benchmarks measured by Vultr on B200 · vLLM. Throughput and latency vary with concurrency, input length and engine configuration, so treat these as a comparison baseline, not a service guarantee.
NVIDIA HGX B200
Unlock Next-Generation AI Power on Vultr Cloud GPU