What GPU do I need to run arcee-ai/Trinity-Large-Preview?
A 399B (MoE) language model for chat and instruction-following. 398.6B parameters, published in BF16. View on Hugging Face
Trinity-Large-Preview is published by arcee-ai on Hugging Face, with 563 downloads and 181 likes to date. It's a AfmoeForCausalLM model built for text-generation, published natively in BF16.
What Trinity-Large-Preview is
Trinity-Large-Preview is a 399B-parameter mixture-of-experts language model published by Arcee AI on Hugging Face. It is released under Custom license.
License note: a lab-specific license (tagged "other" on Hugging Face); read the model card's own license section before commercial use. Facts in this section are sourced from Trinity-Large-Preview's Hugging Face model card, not benchmarked by Aquanode.
What it's used for
- Chat assistants
- Instruction following
- Synthetic data generation
VRAM required & cheapest live GPU fit
Required VRAM = weight size at each precision times a flat 1.2 overhead for activations and fragmentation. The KV-cache grows with context and is not in that factor; it is listed per model below where the architecture is published. Full formula and assumptions: methodology.
| Precision | Weight size | Required VRAM | Cheapest live fit | GPUs needed | Est. $/hr (full fit) |
|---|---|---|---|---|---|
| BF16 | 742.5 GB | 891.0 GB | AMD MI300X | 5 | $13.14/hr |
| FP8 (quantized) | 371.3 GB | 445.5 GB | AMD MI300X | 3 | $7.89/hr |
| INT4 (quantized) | 185.6 GB | 222.8 GB | RTX A6000 | 5 | $1.81/hr |
A GPU is only matched to a row if its hardware supports that precision, and the primary recommendation is always a single-GPU fit when one exists.
INT4 requires a quantized checkpoint actually published for this model, check its Hugging Face page before relying on this row.
Cheapest way to run Trinity-Large-Preview at its published (BF16) precision: 5× AMD MI300X, at $2.63/hr per GPU ($13.14/hr total). Quantizing to FP8 or INT4 (rows above) can cost less, but requires a compatible quantized checkpoint to exist for this model.
Trinity-Large-Preview: KV cache by context length
The KV cache is the memory the attention layers hold for every token of context, on top of the weights. It is not part of the flat 1.2x overhead in the table above, grows with context length and with every concurrent request, and is why a long-context deployment needs more VRAM than the table shows.
Sliding-window attention: windowed layers stop growing at the window, the full-attention layers keep growing. Cached per token: 245,760 bytes at 16-bit while the context is inside the window.
| Context | KV cache, one sequence |
|---|---|
| 4K | 0.94 GB |
| 32K | 2.6 GB |
| 128K | 8.2 GB |
| 256K (model maximum) | 15.7 GB |
Computed from the layer, head and window counts in the model's own configuration (arcee-ai/Trinity-Large-Preview), at 16 bits per cached value. Each concurrent sequence needs its own cache, on top of the weights.
Trinity-Large-Preview: common questions
Can Trinity-Large-Preview run on a single GPU?
No. At BF16 it needs 891.0 GB of VRAM, more than a single 32 GB desktop card holds. The cheapest capable card in the live feed is a 192.0 GB AMD MI300X, and it takes 5 of them.
How many GPUs do I need to run Trinity-Large-Preview?
5 at BF16. It needs 891.0 GB of VRAM and the cheapest capable live offer is a 192.0 GB AMD MI300X, so 5 of them come to $13.14/hr in total.
Does quantizing Trinity-Large-Preview lower the GPU bill?
Yes. At BF16 the cheapest live fit is 5 AMD MI300X cards at $13.14/hr. At INT4 (quantized) it drops to 5 RTX A6000 cards at $1.81/hr, provided a quantized checkpoint exists for it.
How to run Trinity-Large-Preview
Run Trinity-Large-Preview with vLLM
From arcee-ai/Trinity-Large-Preview's own deployment docs.
vllm serve arcee-ai/Trinity-Large-Preview \
--dtype bfloat16 \
--enable-auto-tool-choice \
--tool-call-parser hermesSource: https://huggingface.co/arcee-ai/Trinity-Large-Preview/raw/main/README.md
Run Trinity-Large-Preview with GGUF quantizations
Prebuilt GGUF weights published at unsloth/Trinity-Large-Preview-GGUF. Run with llama.cpp's llama-server or load the repo directly in LM Studio.
llama-server -hf unsloth/Trinity-Large-Preview-GGUFSource: https://huggingface.co/unsloth/Trinity-Large-Preview-GGUF
Deploy Trinity-Large-Preview on Aquanode
Aquanode has no one-click deploy template for Trinity-Large-Preview; you install the inference engine yourself with the commands below. Aquanode sells GPU pods billed per second, not a hosted inference API.
- Launch a bare GPU pod sized to the requirement above (5× AMD MI300X or larger).
- Open a terminal on the pod, or save one of the commands above as a startup script so it runs automatically the first time the pod boots.
- Run the command and connect to the resulting endpoint.
Weight-to-VRAM math, the fit rules, and how live prices are normalized: full methodology.
More Trinity models
- Trinity-Mini (26.1B, BF16)
- Trinity-Nano-Preview (6.1B, BF16)
Alternatives at this size
Other models for text-generation within about a third of Trinity-Large-Preview's 398.6B parameters, from other model lines.
- MiMo-V2.5 (310.8B, F8_E4M3)
- NVIDIA-Nemotron-3-Ultra-550B-A55B-BF16 (560.5B, BF16)
- Ornith-1.5-397B (403.4B, BF16)
- Llama-3.1-405B (405.9B, BF16)
- GLM-4.5 (358.3B, BF16)
More on Trinity-Large-Preview
Fits on a 288 GB GPU at INT4: every model that fits in 288 GB.
Best chat and assistants models: how Trinity-Large-Preview ranks against the rest.
Related reading: AMD MI300X pricing and specs, How much VRAM you need for LLMs, Serving LLMs with vLLM, vLLM vs TensorRT-LLM vs SGLang, and Best GPU for LLM inference.
Your own GPU cloud accounts
One console for the GPU clouds you already rent from. Your keys stay on your machine.
$20 a month, flat