Poro 34B needs about 35.6 GB of memory at Q4_K_M (a 22.4 GB file) with an 8K context window, so it fits on a 48 GB GPU. This repository has 22 quants, from 9.4 GB (IQ3_M) to 37.3 GB (Q8_0).
Poro 34B – 34B‑parameter Bloom decoder for Finnish, English and code
Poro 34B is a 34B‑parameter Bloom decoder-only transformer. It was trained on 1 trillion tokens from Finnish, English and code datasets. The model is released under the Apache‑2.0 license.
Good at
- Large‑scale text generation
- Code completion
- Multilingual support
Limits
- Requires 8 k context tokens
- No vision support
- Large VRAM needed for higher quants
Best for
- Research on open‑source LLMs
- Developing code assistants
- Generating Finnish language content
Who it’s for. Researchers and developers seeking a large, open‑source language model that supports Finnish and English. The model is suitable for those who need a decoder‑only architecture for text and code generation tasks.
Quants & memory
| Quant | File | Est. memory | vs VRAM | Result | |
|---|---|---|---|---|---|
| IQ3_M | 9.4 GB | 21.0 GB | Too big | Get file → | |
| Q2_K | 13.4 GB | 25.0 GB | Too big | Get file → | |
| Q4_K_S | 14.5 GB | 26.1 GB | Too big | Get file → | |
| IQ3_XS | 15.1 GB | 26.7 GB | Too big | Get file → | |
| IQ3_S | 15.5 GB | 27.1 GB | Too big | Get file → | |
| Q3_K_S | 15.5 GB | 27.1 GB | Too big | Get file → | |
| Q3_K | 18.5 GB | 30.1 GB | Too big | Get file → | |
| Q3_K_M | 18.5 GB | 30.1 GB | Too big | Get file → | |
| IQ4_XS | 19.1 GB | 30.7 GB | Too big | Get file → | |
| Q4_0 | 20.0 GB | 31.6 GB | Too big | Get file → | |
| Q3_K_L | 20.2 GB | 31.8 GB | Too big | Get file → | |
| IQ4_NL | 20.2 GB | 31.8 GB | Too big | Get file → | |
| Q4_1 | 22.2 GB | 33.8 GB | Too big | Get file → | |
| Q4_K | 22.4 GB | 34.0 GB | Too big | Get file → | |
| Q4_K_M | 22.4 GB | 34.0 GB | Too big | Get file → | |
| Q5_0 | 24.3 GB | 35.9 GB | Too big | Get file → | |
| Q5_K_S | 24.3 GB | 35.9 GB | Too big | Get file → | |
| Q5_K | 26.1 GB | 37.7 GB | Too big | Get file → | |
| Q5_K_M | 26.1 GB | 37.7 GB | Too big | Get file → | |
| Q5_1 | 26.4 GB | 38.0 GB | Too big | Get file → | |
| Q6_K | 28.8 GB | 40.4 GB | Too big | Get file → | |
| Q8_0 | 37.3 GB | 48.9 GB | Too big | Get file → |
For 8 GB VRAM, use IQ3_M (9.42 GB file, ~22.6 GB memory). For 12 GB VRAM, Q4_K_S (14.53 GB file, ~27.7 GB memory) is suitable. For 24 GB VRAM, Q4_K_M (22.44 GB file, ~35.6 GB memory) provides a balance of size and performance.
What GPU do you need for Poro 34B?
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: no quant fits at an 8K context.
- 16 GB VRAM: no quant fits at an 8K context.
- 24 GB VRAM: the best quant that fits is IQ3_M (9.4 GB file, about 22.6 GB in use).
- 32 GB VRAM: the best quant that fits is Q3_K (18.5 GB file, about 31.7 GB in use).
- 48 GB VRAM: the best quant that fits is Q6_K (28.8 GB file, about 42.0 GB in use).
Which Poro 34B quant should you choose?
Q4_K_M is the usual starting point for Poro 34B: it balances file size and quality well. Here it is a 22.4 GB file.
The Q8_0 file is 37.3 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 IQ3_M at 9.4 GB, with the quality trade-offs that come with fewer bits. A reasonable choice when a 4-bit file will not fit, mostly on larger models. Quality loss is noticeable but often acceptable.
New to quant names? Read GGUF quantization explained.
Specifications
Architecture from the GGUF header
Source
Frequently asked questions about Poro 34B
How much VRAM does Poro 34B need?
About 35.6 GB at Q4_K_M with an 8K context window. That is the 22.4 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 Poro 34B on an 8 GB GPU?
Not at an 8K context. The smallest file needs about 22.6 GB, so you would need a larger GPU or CPU offloading.
What is the file size of Poro 34B GGUF?
It depends on the quant: IQ3_M is 9.4 GB, Q2_K is 13.4 GB, Q4_K_S is 14.5 GB, IQ3_XS is 15.1 GB, IQ3_S is 15.5 GB, Q3_K_S is 15.5 GB and 16 more.
What context length does Poro 34B support?
The GGUF header lists a trained context length of 7,168 tokens (7K). The KV cache grows with the context you use: about 12.7 GB at 8K and 11.1 GB at 7K.
Where can I download Poro 34B GGUF?
The files are on Hugging Face in the RichardErkhov/LumiOpen_-_Poro-34B-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 Poro 34B 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 Poro 34B on 8 GB VRAM?
Use the IQ3_M quant, which is 9.42 GB file and requires about 22.6 GB memory at 8 k context.
Does Poro 34B support code generation?
Yes, it was trained on code datasets and is suitable for code completion tasks.
What license is Poro 34B released under?
Apache‑2.0, fully open source.