PH

Phi 3.5 Mini ChatML

PhiGGUF by mradermacherUpdated 2024-09-12
Download on Hugging Face
Parameters3.82B
Context128K
Layers32
Architecturephi3
Smallest quant1.4 GB
Quick answer

Phi 3.5 Mini ChatML needs about 6.1 GB of memory at Q4_K_M (a 2.4 GB file) with an 8K context window, so it fits on a 8 GB GPU. This repository has 15 quants, from 1.4 GB (Q2_K) to 7.6 GB (F16). Longer context needs more: at 32K tokens the KV cache alone adds about 12.9 GB.

Phi 3.5 Mini ChatML – 3.82B parameter language model

Phi 3.5 Mini ChatML is a 3.82B parameter model based on the phi3 architecture. It was merged from multiple Phi‑3.5 variants and supports a 131072 token context window. The model is available in a variety of static GGUF quantizations.

Good at

  • Chat‑style interactions
  • Short‑form content generation
  • Low‑latency inference

Limits

  • No vision support
  • License not specified
  • Requires 5–11 GB VRAM depending on quant

Best for

  • Building lightweight chatbots
  • Rapid prototyping of LLM apps
  • Running on mid‑range GPUs

Who it’s for. Developers and hobbyists who need a small yet capable language model for local inference. The model is suitable for projects that require a 131072 token context but have limited GPU memory.

Quants & memory

QuantFileEst. memoryvs VRAMResultDownload
Q2_K1.4 GB 5.1 GB
Fits Get file →
IQ3_XS1.6 GB 5.3 GB
Fits Get file →
IQ3_S1.7 GB 5.4 GB
Fits Get file →
Q3_K_S1.7 GB 5.4 GB
Fits Get file →
IQ3_M1.9 GB 5.6 GB
Fits Get file →
Q3_K_M2.0 GB 5.7 GB
Fits Get file →
IQ4_XS2.1 GB 5.8 GB
Fits Get file →
Q3_K_L2.1 GB 5.8 GB
Fits Get file →
Q4_K_S2.2 GB 5.9 GB
Fits Get file →
Q4_K_M2.4 GB 6.1 GB
Fits Get file →
Q5_K_S2.6 GB 6.4 GB
Fits Get file →
Q5_K_M2.8 GB 6.5 GB
Fits Get file →
Q6_K3.1 GB 6.9 GB
Fits Get file →
Q8_04.1 GB 7.8 GB
Tight Get file →
F167.6 GB 11.4 GB
Too big Get file →

For 8 GB VRAM, use Q4_K_M (est. 6.1 GB) or Q4_K_S (5.9 GB). If 12 GB is available, Q5_K_M (6.5 GB) offers higher precision. For the smallest footprint, Q2_K (5.1 GB) is suitable but with lower quality. Estimates are approximate.

What GPU do you need for Phi 3.5 Mini ChatML?

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: no quant fits at an 8K context.
  • 6 GB VRAM: the best quant that fits is Q4_K_S (2.2 GB file, about 5.9 GB in use).
  • 8 GB VRAM: the best quant that fits is Q8_0 (4.1 GB file, about 7.8 GB in use).
  • 12 GB VRAM: the best quant that fits is F16 (7.6 GB file, about 11.4 GB in use).

Which Phi 3.5 Mini ChatML quant should you choose?

Q4_K_M is the usual starting point for Phi 3.5 Mini ChatML: it balances file size and quality well. Here it is a 2.4 GB file.

The Q8_0 file is 4.1 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 Q2_K at 1.4 GB, with the quality trade-offs that come with fewer bits. Use it only when memory is very tight. Expect a clear drop in quality, which is usually worse on small models and easier to live with on large ones.

New to quant names? Read GGUF quantization explained.

Specifications

Architecture from the GGUF header

Architecturephi3
Parameters3.82B
Layers32
Trained context131,072 tokens
KV cache (fp16)384 KB per token

Source

Quantized bymradermacher
FamilyPhi
Downloads (HF)2,475
Last updated2024-09-12

Frequently asked questions about Phi 3.5 Mini ChatML

How much VRAM does Phi 3.5 Mini ChatML need?

About 6.1 GB at Q4_K_M with an 8K context window. That is the 2.4 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 Phi 3.5 Mini ChatML on an 8 GB GPU?

Yes, with Q8_0: the 4.1 GB file needs about 7.8 GB at 8K context. Larger quants in this repository do not fit in 8 GB.

What is the file size of Phi 3.5 Mini ChatML GGUF?

It depends on the quant: Q2_K is 1.4 GB, IQ3_XS is 1.6 GB, IQ3_S is 1.7 GB, Q3_K_S is 1.7 GB, IQ3_M is 1.9 GB, Q3_K_M is 2.0 GB and 9 more.

What context length does Phi 3.5 Mini ChatML support?

The GGUF header lists a trained context length of 131,072 tokens (128K). The KV cache grows with the context you use: about 3.2 GB at 8K and 12.9 GB at 32K.

Where can I download Phi 3.5 Mini ChatML GGUF?

The files are on Hugging Face in the mradermacher/Phi-3.5-Mini-ChatML-GGUF repository. The download buttons on this page open the file on Hugging Face. ggufmodels.com does not host any model files.

Can I run Phi 3.5 Mini ChatML on 8 GB VRAM?

Yes, using Q4_K_M or Q4_K_S quantizations, which require about 6 GB of VRAM at 8 k context.

Does this model support vision tasks?

No, the model has no vision capabilities.

What is the maximum context length?

The model was trained with a 131072 token context window.

Other GGUFs of Phi-3.5-Mini-ChatML

Similar models

Browse all →

More from mradermacher

All by mradermacher →

More Phi models

All Phi →

Other small models

All small models →