TR

Trinity-Nano-Preview

TrinityGGUF by MungertUpdated 2025-12-03
Download on Hugging Face
Parameters6.12B
Context128K
Layers56
Architectureafmoe
Smallest quant0.1 GB
Quick answer

Trinity-Nano-Preview needs about 4.9 GB of memory at Q4_K_M (a 4.0 GB file) with an 8K context window, so it fits on a 6 GB GPU. This repository has 27 quants, from 0.1 GB to 12.3 GB (BF16). Longer context needs more: at 32K tokens the KV cache alone adds about 1.9 GB.

Trinity‑Nano‑Preview: 6B MoE chat model with 1B active parameters

Trinity‑Nano‑Preview is a 6B MoE model with 1B active parameters, trained on 10T tokens. It supports chat with a tuned personality and includes tools capability. The model is released under apache‑2.0 and is experimental.

Good at

  • Chat with tools
  • Math and code reasoning
  • Small‑size MoE performance

Limits

  • May be unstable in some use cases
  • Requires 128k context length
  • No vision support

Best for

  • Chatting in specialized domains
  • Rapid prototyping of MoE models
  • Testing new quantization techniques

Who it’s for. Developers and researchers who want to experiment with a small, chat‑tuned MoE model. The model is suitable for those who can fine‑tune it for specific domains and are comfortable with an experimental release.

Quants & memory

QuantFileEst. memoryvs VRAMResultDownload
—0.1 GB 1.0 GB
Fits Get file →
IQ1_S1.9 GB 2.9 GB
Fits Get file →
IQ2_XXS2.1 GB 3.1 GB
Fits Get file →
IQ1_M2.1 GB 3.1 GB
Fits Get file →
IQ2_XS2.3 GB 3.3 GB
Fits Get file →
IQ2_S2.3 GB 3.3 GB
Fits Get file →
IQ2_M2.4 GB 3.4 GB
Fits Get file →
Q2_K_S2.5 GB 3.4 GB
Fits Get file →
Q2_K_M2.5 GB 3.5 GB
Fits Get file →
IQ3_XXS2.9 GB 3.8 GB
Fits Get file →
IQ3_XS2.9 GB 3.9 GB
Fits Get file →
IQ3_M3.2 GB 4.2 GB
Fits Get file →
Q3_K_S3.2 GB 4.2 GB
Fits Get file →
Q3_K_M3.3 GB 4.2 GB
Fits Get file →
IQ4_XS3.4 GB 4.4 GB
Fits Get file →
IQ4_NL3.5 GB 4.4 GB
Fits Get file →
Q4_03.7 GB 4.6 GB
Fits Get file →
Q4_K_S3.7 GB 4.7 GB
Fits Get file →
Q4_13.9 GB 4.9 GB
Fits Get file →
Q4_K_M4.0 GB 4.9 GB
Fits Get file →
Q5_04.4 GB 5.4 GB
Fits Get file →
Q5_K_M4.6 GB 5.6 GB
Fits Get file →
Q5_14.7 GB 5.7 GB
Fits Get file →
Q6_K_M5.2 GB 6.1 GB
Fits Get file →
Q8_06.5 GB 7.5 GB
Tight Get file →
F167.8 GB 8.7 GB
Too big Get file →
BF1612.3 GB 13.2 GB
Too big Get file →

For 8 GB VRAM, use IQ2_XXS (2.09 GB) or IQ2_XS (2.31 GB) which need about 3.1–3.3 GB memory at 8 k context. For 12 GB VRAM, Q4_K_M (3.95 GB) or Q5_0 (4.39 GB) are good choices, requiring roughly 4.9–5.4 GB memory. Larger models like Q8_0 or F16 need more than 7 GB memory and are not recommended for low‑budget GPUs.

What GPU do you need for Trinity-Nano-Preview?

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 IQ3_XS (2.9 GB file, about 3.9 GB in use).
  • 6 GB VRAM: the best quant that fits is Q5_1 (4.7 GB file, about 5.7 GB in use).
  • 8 GB VRAM: the best quant that fits is Q8_0 (6.5 GB file, about 7.5 GB in use).
  • 12 GB VRAM: the best quant that fits is F16 (7.8 GB file, about 8.7 GB in use).
  • 16 GB VRAM: the best quant that fits is BF16 (12.3 GB file, about 13.2 GB in use).

Which Trinity-Nano-Preview quant should you choose?

Q4_K_M is the usual starting point for Trinity-Nano-Preview: it balances file size and quality well. Here it is a 4.0 GB file.

The Q8_0 file is 6.5 GB, about 1.7 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 0.1 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

Architectureafmoe
Parameters6.12B (approximate, mixture of experts)
Layers56
Trained context131,072 tokens
KV cache (fp16)56 KB per token

Source

Quantized byMungert
FamilyTrinity
Licenseapache-2.0
Downloads (HF)1,040
Last updated2025-12-03

Frequently asked questions about Trinity-Nano-Preview

How much VRAM does Trinity-Nano-Preview need?

About 4.9 GB at Q4_K_M with an 8K context window. That is the 4.0 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-Nano-Preview on an 8 GB GPU?

Yes, with Q8_0: the 6.5 GB file needs about 7.5 GB at 8K context. Larger quants in this repository do not fit in 8 GB.

What is the file size of Trinity-Nano-Preview GGUF?

It depends on the quant: file is 0.1 GB, IQ1_S is 1.9 GB, IQ2_XXS is 2.1 GB, IQ1_M is 2.1 GB, IQ2_XS is 2.3 GB, IQ2_S is 2.3 GB and 21 more.

What context length does Trinity-Nano-Preview support?

The GGUF header lists a trained context length of 131,072 tokens (128K). The KV cache grows with the context you use: about 0.5 GB at 8K and 1.9 GB at 32K.

Where can I download Trinity-Nano-Preview GGUF?

The files are on Hugging Face in the Mungert/Trinity-Nano-Preview-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-Nano-Preview 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‑Nano‑Preview on 8 GB VRAM?

Yes, use a quant such as IQ2_XXS or IQ2_XS, which are under 3.5 GB memory at 8 k context.

Does this model support vision?

No, the model has no vision capability.

What license does it use?

The model is released under the apache‑2.0 license.

Other GGUFs of Trinity-Nano-Base

Similar models

Browse all →

More from Mungert

All by Mungert →

More Trinity models

All Trinity →

Other small models

All small models →