AL

ALIA 40b

AliaGGUF by mradermacherUpdated 2025-01-22
Download on Hugging Face
Parameters40.4B
Context4K
Layers48
Architecturellama
Smallest quant9.7 GB
Quick answer

ALIA 40b needs about 26.7 GB of memory at Q4_K_M (a 24.6 GB file) with an 8K context window, so it fits on a 32 GB GPU. This repository has 23 quants, from 9.7 GB (IQ1_S) to 33.2 GB (Q6_K).

40B‑parameter llama‑style model focused on coding

ALIA 40b is a 40.4B parameter decoder‑only language model. It was pre‑trained on 9.37 trillion tokens in 35 European languages and code. The base model is intended for research and commercial use and can be fine‑tuned for specific tasks.

Good at

  • Large‑scale code generation
  • Multilingual text understanding
  • Research prototyping

Limits

  • No instruction tuning or alignment built in
  • Requires high VRAM for full context
  • No vision or multimodal support

Best for

  • Fine‑tuning for code assistants
  • Large‑scale research experiments
  • Custom language generation pipelines

Who it’s for. Researchers and developers who need a large, multilingual base model for code and text. The model is suitable for those who plan to add instruction tuning or alignment before deployment.

Quants & memory

QuantFileEst. memoryvs VRAMResultDownload
IQ1_S9.7 GB 11.0 GB
Too big Get file →
IQ1_M10.5 GB 11.8 GB
Too big Get file →
IQ2_XXS11.7 GB 13.0 GB
Too big Get file →
IQ2_XS12.8 GB 14.1 GB
Too big Get file →
IQ2_S13.6 GB 14.9 GB
Too big Get file →
IQ2_M14.5 GB 15.8 GB
Too big Get file →
Q2_K_S14.7 GB 16.0 GB
Too big Get file →
Q2_K15.7 GB 17.0 GB
Too big Get file →
IQ3_XXS16.2 GB 17.5 GB
Too big Get file →
IQ3_XS17.4 GB 18.7 GB
Too big Get file →
Q3_K_S18.2 GB 19.5 GB
Too big Get file →
IQ3_S18.3 GB 19.6 GB
Too big Get file →
IQ3_M18.8 GB 20.1 GB
Too big Get file →
Q3_K_M20.0 GB 21.4 GB
Too big Get file →
Q3_K_L21.6 GB 22.9 GB
Too big Get file →
IQ4_XS22.2 GB 23.5 GB
Too big Get file →
Q4_023.4 GB 24.7 GB
Too big Get file →
Q4_K_S23.4 GB 24.8 GB
Too big Get file →
Q4_K_M24.6 GB 25.9 GB
Too big Get file →
Q4_125.7 GB 27.0 GB
Too big Get file →
Q5_K_S28.1 GB 29.4 GB
Too big Get file →
Q5_K_M28.8 GB 30.1 GB
Too big Get file →
Q6_K33.2 GB 34.5 GB
Too big Get file →

For 8 GB VRAM, use IQ1_S (9.73 GB file, ~11.8 GB memory). For 12 GB VRAM, IQ2_M (14.54 GB file, ~16.7 GB memory) is a good balance. For 16 GB VRAM, Q3_K_M (20.04 GB file, ~22.2 GB memory) offers higher quality.

What GPU do you need for ALIA 40b?

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_S (9.7 GB file, about 11.8 GB in use).
  • 16 GB VRAM: the best quant that fits is IQ2_S (13.6 GB file, about 15.7 GB in use).
  • 24 GB VRAM: the best quant that fits is Q3_K_L (21.6 GB file, about 23.7 GB in use).
  • 32 GB VRAM: the best quant that fits is Q5_K_M (28.8 GB file, about 30.9 GB in use).
  • 48 GB VRAM: the best quant that fits is Q6_K (33.2 GB file, about 35.3 GB in use).

Which ALIA 40b quant should you choose?

Q4_K_M is the usual starting point for ALIA 40b: it balances file size and quality well. Here it is a 24.6 GB file.

If memory is tight, the smallest file is IQ1_S at 9.7 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

Architecturellama
Parameters40.4B
Layers48
Trained context4,096 tokens
KV cache (fp16)192 KB per token

Source

Quantized bymradermacher
FamilyAlia
Base modelBSC-LT/ALIA-40b
Licenseapache-2.0
Downloads (HF)986
Last updated2025-01-22

Frequently asked questions about ALIA 40b

How much VRAM does ALIA 40b need?

About 26.7 GB at Q4_K_M with an 8K context window. That is the 24.6 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 ALIA 40b on an 8 GB GPU?

Not at an 8K context. The smallest file needs about 11.8 GB, so you would need a larger GPU or CPU offloading.

What is the file size of ALIA 40b GGUF?

It depends on the quant: IQ1_S is 9.7 GB, IQ1_M is 10.5 GB, IQ2_XXS is 11.7 GB, IQ2_XS is 12.8 GB, IQ2_S is 13.6 GB, IQ2_M is 14.5 GB and 17 more.

What context length does ALIA 40b support?

The GGUF header lists a trained context length of 4,096 tokens (4K). The KV cache grows with the context you use: about 1.6 GB at 8K and 0.8 GB at 4K.

Where can I download ALIA 40b GGUF?

The files are on Hugging Face in the mradermacher/ALIA-40b-i1-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 ALIA 40b 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 ALIA 40b on 8 GB VRAM?

Use the IQ1_S quant, which requires about 11.8 GB of VRAM at 8k context.

Does this model support vision tasks?

No, the model is a text‑only decoder and has no vision capability.

Is the model aligned with user instructions?

No, the base model lacks instruction tuning; alignment should be added before production use.

Other GGUFs of ALIA-40b

Similar models

Browse all →

More from mradermacher

All by mradermacher →

More Alia models

All Alia →

Other large models

All large models →