Trinity Large Thinking
arcee-trinity-large-thinkingtrinity-large-thinking- 🧠 Reasoning-optimized variant of Arcee AI's Trinity-Large 398B sparse MoE family.
- ⚡ Roughly 13B active parameters per token for efficient inference.
- 📏 256K context window for long, multi-step agentic chains.
- 💬 Emits extended chain-of-thought inside reasoning-trace blocks.
- 🔧 Tool calling and agentic RL post-training for long-horizon tasks.
- 🌐 Multilingual training across 14 non-English languages.
- 🔒 Released under Apache 2.0 in 2026.
- 🏢 Built on Trinity-Large-Base; trained with Muon optimizer and SMEBU.
Arcee AI is an artificial intelligence company focused on developing advanced language models. The organization has built a reputation in the open-source AI community for its work on model optimization and specialized text generation architectures.
Trinity Large Thinking is the reasoning-oriented member of Arcee AI's Trinity-Large series, a sparse Mixture-of-Experts model with roughly 398–400B total parameters and about 13B activated per token. It shares the same MoE architecture as the chat-focused Trinity-Large-Preview but is post-trained for extended chain-of-thought reasoning and agentic reinforcement learning, making it suited to long-horizon agents, multi-turn tool calling, and audit-friendly stepwise output.
The chief distinction from its same-family predecessors is reasoning behavior. Where Trinity-Large-Preview is lightly post-trained and chat-ready without trace output, Thinking emits intermediate reasoning inside dedicated reasoning-trace blocks before its final answer, and it is built on the Trinity-Large-Base foundation rather than being a fresh pretraining run.
Architecturally, the Trinity-Large family uses 256 experts with 4 active per token, interleaved local and global attention, gated attention, and sigmoid routing, according to Arcee's technical report. Training used the Muon optimizer plus a load-balancing technique called Soft-clamped Momentum Expert Bias Updates (SMEBU) across a 17-trillion-token pretraining recipe, completing with zero loss spikes.
The model supports tool calling, multilingual input, and a large context window for sustained agentic workflows. It is distributed under Apache 2.0, with FP8 weights and quantized GGUF builds available for self-hosting.
This About section is AI-generated from public sources via VeniceStats + Venice inference, with no human editing. It may contain inaccuracies.
Daily usage · Trinity Large Thinking · Organic sellers
| Seller | Rep↓ | Sales | Tokens | Routing | Input $/M | Cached $/M | Output $/M | Save |
|---|---|---|---|---|---|---|---|---|
| Apex TEE Test 0xe672…7955 | 0.00 | — | — | gated | $0.2438 | $0.0585 | $0.8775 | −22% |
| antseed-neon-puma-944e 0x6650…944e | 0.00 | — | — | gated | $0.1066 | $0.075 | $0.3836 | −66% |
| Meridian AI 0x8c8c…06f5 | 0.00 | — | — | gated | $0.0844 | $0.0844not advertised | $0.3038 | −73% |
| Leftermute 0x388b…5389 | 0.00 | — | — | gated | $0.0631 | $0.0631not advertised | $0.2273 | −80% |
"Best price" and the seller table are live Antseed catalog data (advertised $/1M tokens — or $ per generated image for unit-billed image models — not settled amounts). Reputation = buyer trust score (0-100, the Antseed SDK's own formula). "Routing" = the SDK's default buyer routing (what the VPR desktop app ships with): a trust ≥ 60 gate on the effective reputation, then cheapest-first among routable sellers; live failover state (per-peer cooldowns) is buyer-side runtime and not included. Model knowledge (TLDR, provider, About) via the VeniceStats enrichment layer. Advertised catalog, not the model used in any specific purchase. "Usage on Antseed" counts only settlements whose buyers share the per-model split on-chain (metadata v2/v3, opt-in), so every usage figure is a lower bound.