Open models by OpenAI
Previously listed as free · not rechecked
Open-weight reasoning models you can run, customize, and connect to your own tools instead of sending every task to a hosted API.
AI Agents & Chatbots
AI Agents & Chatbots Checked Oct 19, 2026
Trinity Large Thinking
Evaluate a large open reasoning model for planning, tool calling, and agent workflows.
Best for: Developers and researchers evaluating open models for structured reasoning or tool-connected systems.
Access
Previously listed as Trinity Large Thinking is released under the OpenMDW-1.1 license with downloadable weights for eligible local or hosted inference. The model is a roughly 398B-parameter sparse MoE with about 13B active parameters per token and an extended context path, so serving it needs serious infrastructure. Hosted chat or API access has separate limits and pricing.
Useful?
What you can do with it
- Test multi-step reasoning
- Build a tool-calling agent
- Evaluate an open model against a hosted assistant
- Prototype a research workflow
- Serve a reasoning model with an inference stack
Good to know
- Current open-weight model card
- Reasoning and tool-calling focus
- Sparse MoE architecture
- Extended context path
Limitations
- Very large infrastructure requirement
- Hosted access is separate from weights
- License and deployment terms need review
Other official links
Keep exploring
Similar tools
OpenManus
Previously listed as free · not rechecked
Build and inspect an agent that can plan, browse, use tools, and work through multi-step tasks.
AI Agents & Chatbots
Qwen Studio
Previously listed as free tier · not rechecked
Use Qwen Studio for free chat, reasoning, coding, web research, and multimodal creation.
AI Agents & Chatbots