Entity trail

Arcee AI

Source-backed findings, relationship evidence, citations, and briefing history from the public MindPattern archive.

Briefing refs

Findings

Edges

Sources

Corpus findings

2026-04-12 / projects-researcherArcee AI Releases Trinity Large Thinking: Most Capable Non-Chinese Open-Weight ModelArcee AI, a small open-source AI model maker, released Trinity Large Thinking which it claims is the most capable open-weight model ever released by a non-Chinese company. TechCrunch covered the release on April 7, highlighting Arcee's bootstrapped approach to competing with much larger labs. For builders evaluating open-weight alternatives to proprietary APIs, Trinity represents another option in the increasingly competitive open model space.
2026-04-02 / sources-researcherLatent Space AINews: Quiet April Fools Recap — Arcee, Z.ai, TII, Claude Code Leak Aftermath, Multi-Agent ResearchLatent Space's April 2 AINews roundup aggregates the day's key signals: Arcee Trinity-Large-Thinking launch, Z.ai GLM-5V-Turbo release, TII Falcon Perception debut, Claude Code 512K-line source exposure aftermath (Anthropic DMCA takedowns), and the self-organizing multi-agent study. Also notes Liquid AI received 'best April Fools joke' recognition and OpenAI Codex rate limits reset across all plans. Valuable as a cross-reference for triangulating which stories the AI engineering community considers most significant.
2026-04-02 / sources-researcherArcee Trinity-Large-Thinking: 400B Open-Source MoE Hits #2 on PinchBench at 96% Less Than Opus 4.6Arcee AI released Trinity-Large-Thinking under Apache 2.0 — a 400B sparse MoE with 13B active parameters (256 experts, 4 active per token) optimized for complex long-horizon agents and multi-turn tool calling. It ranks #2 on PinchBench behind only Opus 4.6 while costing ~$0.90/M output tokens (vs Opus pricing), and is described as the strongest open model ever released outside of China. Significant improvements over Preview in multi-turn tool use, context coherence, and instruction following across extended agent runs.
2026-02-23 / sources-researcherArcee Trinity Large — Open 400B Sparse MoE 13B ActiveLargest open-weight US model. 400B total, 13B active. 87.2 MMLU. Apache 2.0. Trained in 33 days on 2048 B300s for ~20M.

Graph relationships

RELEASED
Arcee AI -> Trinity Large Thinking
Arcee AI released Trinity Large Thinking, claimed to be the most capable open-weight model by a non-Chinese company.
Source finding
Reported in
Arcee Trinity-Large-Thinking: 400B Open-Source MoE Hits #2 on PinchBench at 96% Less Than Opus 4.6
Source finding

Source trail

Vellum AI — Opus 4.6 Benchmarks Arcee AI — Trinity Large arXiv — PCAS arXiv — SpargeAttention2 arXiv — CUWM arXiv — VESPO arXiv — SAGE The Register — TDD for AI Agents

Graph sources

entity graphfindings textkg edgeskg entitiesnewsletter issues