POCKET 35B needs about 22.3 GB of memory at Q4_K_M (a 21.2 GB file) with an 8K context window, so it fits on a 24 GB GPU. This repository has 4 quants, from 8.2 GB (IQ1_M) to 21.2 GB (Q4_K_M). Longer context needs more: at 32K tokens the KV cache alone adds about 2.7 GB.
35B on-device LLM that runs on CPU or phone
POCKET 35B is a 35‑billion‑parameter mixture‑of‑experts model based on Darwin‑36B‑Opus. It is quantized for on‑device use and runs with stock llama.cpp on CPUs and phones without GPU. The model supports a 262,144‑token context and is released under Apache‑2.0.
Good at
- Large context handling
- On‑device CPU inference
- No GPU requirement
Limits
- Requires 8‑GB+ VRAM for larger quants
- No vision support
- License is Apache‑2.0
Best for
- Chatting on phones
- CPU‑only desktop inference
- Low‑cost deployment
Who it’s for. Developers and hobbyists who need a large language model that can run on a laptop or phone without a GPU. The model is suitable for applications that require a long context window and can tolerate the memory demands of the chosen quantization.
Quants & memory
| Quant | File | Est. memory | vs VRAM | Result | |
|---|---|---|---|---|---|
| IQ1_M | 8.2 GB | 9.4 GB | Too big | Get file → | |
| Q2_K | 12.9 GB | 14.1 GB | Too big | Get file → | |
| Q3_K_M | 16.8 GB | 17.9 GB | Too big | Get file → | |
| Q4_K_M | 21.2 GB | 22.3 GB | Too big | Get file → |
For 8‑GB VRAM use IQ1_M (8.24 GB file, ~9.4 GB RAM). With 12‑GB VRAM Q2_K (12.94 GB file, ~14.1 GB RAM) is suitable. For 16‑GB VRAM or more, Q3_K_M (16.76 GB file, ~17.9 GB RAM) or Q4_K_M (21.17 GB file, ~22.3 GB RAM) can be used. Memory estimates are approximate.
What GPU do you need for POCKET 35B?
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: no quant fits at an 8K context.
- 8 GB VRAM: no quant fits at an 8K context.
- 12 GB VRAM: the best quant that fits is IQ1_M (8.2 GB file, about 9.4 GB in use).
- 16 GB VRAM: the best quant that fits is Q2_K (12.9 GB file, about 14.1 GB in use).
- 24 GB VRAM: the best quant that fits is Q4_K_M (21.2 GB file, about 22.3 GB in use).
Which POCKET 35B quant should you choose?
Q4_K_M is the usual starting point for POCKET 35B: it balances file size and quality well. Here it is a 21.2 GB file.
If memory is tight, the smallest file is IQ1_M at 8.2 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
Source
Frequently asked questions about POCKET 35B
How much VRAM does POCKET 35B need?
About 22.3 GB at Q4_K_M with an 8K context window. That is the 21.2 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 POCKET 35B on an 8 GB GPU?
Not at an 8K context. The smallest file needs about 9.4 GB, so you would need a larger GPU or CPU offloading.
What is the file size of POCKET 35B GGUF?
It depends on the quant: IQ1_M is 8.2 GB, Q2_K is 12.9 GB, Q3_K_M is 16.8 GB, Q4_K_M is 21.2 GB.
What context length does POCKET 35B support?
The GGUF header lists a trained context length of 262,144 tokens (256K). The KV cache grows with the context you use: about 0.7 GB at 8K and 2.7 GB at 32K.
Where can I download POCKET 35B GGUF?
The files are on Hugging Face in the FINAL-Bench/POCKET-35B-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 POCKET 35B 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 POCKET 35B on a laptop with 8 GB VRAM?
Yes, the IQ1_M quantization uses a 8.24 GB file and requires about 9.4 GB of RAM at 8k context, which fits on a laptop with 8 GB VRAM and sufficient system RAM.
Does POCKET 35B support image inputs?
No, the model has no vision capability; it is a text‑only language model.