The first entry is always the pin's source. Overall confidence is a weighted average of how firmly each reference supports the start and end times used above; a reference counts half as much for every 180 days older than the newest.
Arcee's[2] post "Trinity-Large-Thinking: Scaling an Open Source Frontier Agent" is dated Apr 1, 2026 and says "Today we are releasing Trinity-Large-Thinking on our API and the weights on Hugging Face[4]"; OpenRouter[5] lists it as released Apr 1, 2026.
Arcee's[1] later post moving the whole Trinity family, including Trinity-Large-Thinking, from Apache 2.0 to the Linux Foundation's OpenMDW-1.1 licence.
Secondary write-up dated April 2, 2026 confirming the April 1 release, the 398B/13B sparsity, 33-day 2,048-B300 run and think-block handling.[1]
Model card: about 398B parameters with about 13B active, 256 experts with 4 active, 17T pretraining tokens on 2,048 B300 GPUs, context extended to 512k, think-block reasoning that must be kept in context, and scores including PinchBench 91.9, AIME25 96.3 and SWE-bench Verified 63.2.[1]
Router listing: released Apr 1, 2026, 262K context and $0.25 per 1 million input tokens and $0.80 per 1 million output tokens.[1]
Something missing or wrong? Say it in your own words: a link that backs this pin up, a different start or end date and why, or a fact it lacks or gets wrong. The AI checks it against this pin's sources, searches for better ones, and adds any page that backs you up. The pin's own sources still count most. A picture that shows something else, or shows it badly, is looked at too, and moved down or replaced.