Clarifai selected as inference provider for Arcee AI's Trinity open-weight models

WASHINGTON, Dec. 2, 2025 /PRNewswire/ -- Clarifai, a global leader in AI and pioneer of the full-stack AI platform, today announced a new partnership with Arcee AI to serve as the primary launch partner and hosting provider for Trinity, Arcee's new family of open-weight foundation models. The collaboration not only accelerates Arcee's launch but also demonstrates a repeatable blueprint for other model labs looking to bring their home-trained or open-weight models to market quickly and efficiently.

The Trinity models—Trinity Nano (6B) and Trinity Mini (26B)—are now fully available through Clarifai's Compute Orchestration platform, giving developers, startups, and enterprises immediate access to high-efficiency, U.S.-built open-weight LLMs served through an OpenAI-compatible API. A larger third Trinity model is currently in development.

This marks a major milestone in the open-weight ecosystem: a U.S. model lab focused on performance-per-parameter partnering with a U.S. infrastructure platform known for delivering high-throughput, low-cost inference at scale.

"Arcee AI is demonstrating what's possible with efficient, open-weight U.S. models, and we're proud to bring their Trinity family to production through Clarifai's Compute Orchestration platform. This launch reflects exactly what our infrastructure is designed for—making it fast, reliable, and cost-effective for model builders to deploy their own custom or open-weight LLMs without having to stand up their own inference stack," said Matt Zeiler, Founder and CEO, Clarifai. "As more teams develop specialized models, Clarifai will continue to be the platform with the best price-performance, that brings those models from research environments to real-world scale."

Bringing Arcee's Efficient, Open-Weight Models to Production

Arcee AI, an American open-intelligence lab, is committed to accelerating the frontier of open-weight models in the U.S. The newly announced Trinity models reflect this mission:

• Trinity Nano (6B / 1B active) — optimized for real-time, on-device, and embedded AI
• Trinity Mini (26B / 3B active) — tuned for multi-step agent workflows, robust function calling, and long-context use cases

Both models share:

• 128K context windows for extended multi-turn reasoning
• Efficient attention mechanisms to reduce memory and compute loads
• High inference efficiency with strong price-performance characteristics

"With the launch of Trinity, we're showing that frontier models can be built quickly, with efficient capital, and still perform at the highest level," said Mark McQuade, CEO, Arcee AI. "In just six months, we've trained three open-weight models with each release increasing in effectiveness. By partnering with Clarifai, the leader in compute orchestration, the pace of collaborative research will continue to accelerate and help us create the next great frontier and foundational models."

Production-Grade Inference, Lower Costs

By launching Trinity on Clarifai, Arcee AI gains a hardened, cost-efficient inference layer purpose-built for custom and open-weight models.

Clarifai's Compute Orchestration platform provides:

• Autoscaling, GPU fractioning, and spot-aware scheduling for lower operational costs
• Kernel-level performance optimizations for high throughput and low latency
• Hybrid, multi-cloud, and on-prem flexibility
• Global availability with enterprise-grade reliability
• OpenAI-compatible endpoints for drop-in integration

Arcee AI's Trinity Nano and Trinity Mini models are available now on Clarifai, with full Playground access, API integration, and immediate compatibility with OpenAI-spec clients, agentic frameworks, and MCP tools.

Developers can get started at https://clarifai.com/.

Learn more at https://arcee.ai.