Arcee Trinity-Large-Thinking Pricing
All plans and pricing as of 2026-04-17
Self-hosted (Apache 2.0)
- ✓Trained from scratch, not a fine-tune of an existing model
- ✓Apache 2.0 license, unrestricted commercial use
- ✓Weights on Hugging Face
- ✓256-expert sparse MoE with 4 experts active (~1.56% routing)
API (OpenRouter, Trinity-Large-Thinking)
- ✓Available on OpenRouter for hosted inference
- ✓~96% cheaper than Claude Opus 4.6 at the same quality tier on agentic tasks
- ✓Pay-as-you-go
Is Arcee Trinity-Large-Thinking Worth the Price?
Value Score: 9.5/10
Overall Score: 8.1/10 · Teams that need a US-made, Apache 2.0, frontier-tier open-weight model and can either rent multi-GPU infrastructure or pay OpenRouter API pricing at ~$0.90/M output tokens. Particularly valuable for US government, defense, or regulated enterprise contexts where country-of-origin matters for procurement. Also good for agentic reasoning workloads where the ~96% cost savings vs Claude Opus actually changes what you can build.
Arcee Trinity-Large-Thinking is the most consequential US-made open-weight launch since Meta's Llama 4. A tiny US startup shipping a 398B-parameter sparse-MoE frontier reasoning model, trained from scratch, under Apache 2.0, priced ~96% below Claude Opus -- that is genuinely a new category of competitor in the open-weight ecosystem. The third-party benchmark verification is still landing, so treat the 'Opus-tier' positioning as provisional through April 2026. But even if Trinity lands at 80% of the claimed quality, it is the strongest US-made open-weight frontier option available today, and for US procurement / country-of-origin-sensitive deployments it fills a real gap that nobody else had solved.
How Arcee Trinity-Large-Thinking Pricing Compares
| Tool | Free Tier | Starting Price | Value Score | Overall |
|---|---|---|---|---|
| Arcee Trinity-Large-Thinking(this tool) | Yes | $0 | 9.5/10 | 8.1 |
| Qwen (Alibaba) | Yes | $0 | 10/10 | 8.8 |
| MiniMax M2 / M2.5 | Yes | $0 | 9.5/10 | 8.4 |
| Gemma 4 (Google) | Yes | $0 | 10/10 | 8.3 |
| IBM Granite 4.0 | Yes | $0 | 9.5/10 | 8.2 |
| Kimi K2.5 (Moonshot) | Yes | $0 | 8.5/10 | 8.1 |