Trinity-Large-Preview is a frontier-scale open-weight language model from Arcee, built as a 400B-parameter sparse Mixture-of-Experts with 13B active parameters per token using 4-of-256 expert routing. It excels in creative writing,...
Tested on 1 benchmarks. Top scores: Chatbot Arena Elo · Overall (1375.1%).
Chatbot Arena overall Elo rating. Crowdsourced human preference ranking from blind head-to-head comparisons across all topics.
- Typetext
- Context131K tokens (~66 books)
- ReleasedJan 2026
- LicenseOpen Source
- Statusexpired
- Cost / Message~$0.000
Frequently Asked Questions
Related Models
ERNIE-5.5-Flash-Preview (free) · baiduQwen3 Coder Next FP8 · AlibabaQwen2.5 Coder 32B Instruct AWQ · AlibabaSpeaker Diarization Community 1 · PyannoteJais 3 0.9B · MBZUAIBenchmarks
Chatbot Arena Elo · OverallIdentifiers and sizes
- Hugging Face repo · arcee-ai/Trinity-Large-Preview
- Total parameters · 399B (398,635,286,016) · source Hugging Face safetensors · fetched 2026-11-14
- Weights as published · 797 GB in BF16 (computed from the safetensors tensor counts per dtype)
- Weights size, computed · 797 GB at BF16/FP16 · 399 GB at INT8/Q8 · 199 GB at Q4 (computed from the parameter count, not measured; excludes KV cache and runtime overhead)
Key facts · as of 2026-04-22
- Trinity Large Preview (free) by arcee-ai. Not enough public benchmark scores to rank yet.
- List price free input · free output per 1M tokens (as of 2026-04-22).
How to cite · data as of 2026-04-22
Trinity Large Preview (free) · benchmarks, pricing and providers. BenchGecko, data as of 2026-04-22. https://benchgecko.ai/model/trinity-large-preview-free
Credit "Source: BenchGecko" with a link. Prices per provider and Gecko Tests are BenchGecko data (CC BY 4.0); benchmark scores keep their original source, listed in the JSON. JSON · llms.txt · MCP