Trinity needs about 244.4 GB of memory at Q4_K_M (a 241.8 GB file) with an 8K context window, which is more than a 48 GB GPU holds. This repository has 29 quants, from 0.5 GB to 423.7 GB (Q8_0). Longer context needs more: at 32K tokens the KV cache alone adds about 8.1 GB.
Trinity-Large-Thinking – a 399B sparse MoE model optimized for reasoning and tool use
Trinity-Large-Thinking is a 399B‑parameter sparse Mixture‑of‑Experts model with extended chain‑of‑thought reasoning. It generates explicit reasoning traces wrapped in ` ... ` blocks before producing a final answer. The model is designed for agentic workflows and tool calling.
Good at
- Reasoning and chain‑of‑thought
- Tool calling and agent workflows
- Multi‑step planning
Limits
- Requires large VRAM for full quantizations
- Thinking tokens must stay in context
- No vision support
Best for
- Agentic chatbots
- Tool‑enabled assistants
- Complex planning tasks
Who it’s for. Developers building agentic applications that need explicit reasoning traces. Researchers exploring sparse MoE architectures and extended chain‑of‑thought.
Quants & memory
| Quant | File | Est. memory | vs VRAM | Result | |
|---|---|---|---|---|---|
| — | 0.5 GB | 3.0 GB | Fits | Get file → | |
| IQ1_S | 81.2 GB (3 parts) | 83.7 GB | Too big | Get file → | |
| IQ1_M | 90.9 GB (3 parts) | 93.4 GB | Too big | Get file → | |
| IQ2_XXS | 106.0 GB (3 parts) | 108.5 GB | Too big | Get file → | |
| IQ2_XS | 118.2 GB (3 parts) | 120.8 GB | Too big | Get file → | |
| IQ2_S | 120.3 GB (4 parts) | 122.8 GB | Too big | Get file → | |
| IQ2_M | 133.0 GB (4 parts) | 135.5 GB | Too big | Get file → | |
| Q2_K | 139.2 GB (4 parts) | 141.7 GB | Too big | Get file → | |
| Q2_K_L | 139.8 GB (4 parts) | 142.3 GB | Too big | Get file → | |
| IQ3_XXS | 165.6 GB (5 parts) | 168.1 GB | Too big | Get file → | |
| Q3_K_S | 173.0 GB (5 parts) | 175.5 GB | Too big | Get file → | |
| IQ3_XS | 181.2 GB (5 parts) | 183.7 GB | Too big | Get file → | |
| Q3_K_M | 181.4 GB (5 parts) | 183.9 GB | Too big | Get file → | |
| Q3_K_L | 189.1 GB (5 parts) | 191.6 GB | Too big | Get file → | |
| IQ3_M | 189.2 GB (5 parts) | 191.8 GB | Too big | Get file → | |
| Q3_K_XL | 189.6 GB (5 parts) | 192.1 GB | Too big | Get file → | |
| IQ4_XS | 213.0 GB (6 parts) | 215.6 GB | Too big | Get file → | |
| IQ4_NL | 225.4 GB (6 parts) | 227.9 GB | Too big | Get file → | |
| Q4_0 | 226.3 GB (6 parts) | 228.8 GB | Too big | Get file → | |
| Q4_K_S | 233.2 GB (6 parts) | 235.7 GB | Too big | Get file → | |
| Q4_K_M | 241.8 GB (7 parts) | 244.4 GB | Too big | Get file → | |
| Q4_K_L | 242.3 GB (7 parts) | 244.8 GB | Too big | Get file → | |
| Q4_1 | 250.1 GB (7 parts) | 252.6 GB | Too big | Get file → | |
| Q5_K_S | 274.9 GB (8 parts) | 277.4 GB | Too big | Get file → | |
| Q5_K_M | 283.6 GB (8 parts) | 286.1 GB | Too big | Get file → | |
| Q5_K_L | 284.0 GB (8 parts) | 286.5 GB | Too big | Get file → | |
| Q6_K | 343.2 GB (9 parts) | 345.7 GB | Too big | Get file → | |
| Q6_K_L | 343.5 GB (9 parts) | 346.0 GB | Too big | Get file → | |
| Q8_0 | 423.7 GB (11 parts) | 426.2 GB | Too big | Get file → |
For 8‑GB VRAM, use Q4_0 (file size 226.27 GB, memory 228.8 GB at 8k context). For 16‑GB VRAM, Q4_K_M (241.85 GB, 244.4 GB). For 24‑GB VRAM, Q5_K_M (283.61 GB, 286.1 GB). Estimates are approximate.
What GPU do you need for Trinity?
These figures add the file size, an fp16 KV cache at an 8K context window (worked out from this model’s own header) and about 0.5 GB for the runtime. They are estimates, so leave a little headroom, and expect to use less context or a smaller quant if you are right at the limit.
- 4 GB VRAM: the best quant that fits is the main file (0.5 GB file, about 3.0 GB in use).
- 6 GB VRAM: the best quant that fits is the main file (0.5 GB file, about 3.0 GB in use).
- 8 GB VRAM: the best quant that fits is the main file (0.5 GB file, about 3.0 GB in use).
- 12 GB VRAM: the best quant that fits is the main file (0.5 GB file, about 3.0 GB in use).
- 16 GB VRAM: the best quant that fits is the main file (0.5 GB file, about 3.0 GB in use).
- 24 GB VRAM: the best quant that fits is the main file (0.5 GB file, about 3.0 GB in use).
- 32 GB VRAM: the best quant that fits is the main file (0.5 GB file, about 3.0 GB in use).
- 48 GB VRAM: the best quant that fits is the main file (0.5 GB file, about 3.0 GB in use).
Which Trinity quant should you choose?
Q4_K_M is the usual starting point for Trinity: it balances file size and quality well. Here it is a 241.8 GB file.
The Q8_0 file is 423.7 GB, about 1.8 times the size of Q4_K_M. It keeps the weights closer to the original, which only matters if you have the memory to spare.
If memory is tight, the smallest file is the first one in the table at 0.5 GB, with the quality trade-offs that come with fewer bits.
New to quant names? Read GGUF quantization explained.
Specifications
Architecture from the GGUF header
Source
Frequently asked questions about Trinity
How much VRAM does Trinity need?
About 244.4 GB at Q4_K_M with an 8K context window. That is the 241.8 GB file, plus the KV cache and roughly 0.5 GB of runtime overhead. Smaller quants need less and longer context needs more.
Can I run Trinity on an 8 GB GPU?
Yes, with the smaller files: the 0.5 GB file needs about 3.0 GB at 8K context. Larger quants in this repository do not fit in 8 GB.
What is the file size of Trinity GGUF?
It depends on the quant: file is 0.5 GB, IQ1_S is 81.2 GB, IQ1_M is 90.9 GB, IQ2_XXS is 106.0 GB, IQ2_XS is 118.2 GB, IQ2_S is 120.3 GB and 23 more.
What context length does Trinity support?
The GGUF header lists a trained context length of 262,144 tokens (256K). The KV cache grows with the context you use: about 2.0 GB at 8K and 8.1 GB at 32K.
Where can I download Trinity GGUF?
The files are on Hugging Face in the bartowski/arcee-ai_Trinity-Large-Thinking-GGUF repository. The download buttons on this page open the file on Hugging Face. ggufmodels.com does not host any model files.
What license does Trinity use?
Hugging Face lists the license as apache-2.0. Licenses can differ between the quantized files and the original model, so check the repository and the base model before commercial use.
Can I run Trinity on 8 GB VRAM?
Yes, using the Q4_0 quantization which requires about 228.8 GB of VRAM at 8k context, but this is a theoretical estimate; actual usage may vary.
Does Trinity support vision tasks?
No, the model has no vision capabilities.