Arcee AI released Trinity-Large-Thinking on April 1 — a 400B sparse MoE reasoning model under Apache 2.0 that activates only 13B parameters per token. It ranks #2 on PinchBench (91.9 vs. Claude Opus 4.6's 93.3) while priced at $0.90 per million output tokens, roughly 96% cheaper than Opus 4.6's $25/M. VentureBeat notes it is built specifically for long-horizon agent loops and multi-turn tool calling rather than pure benchmark performance, supporting a 262K-token context window.