CE

Cerebrum 1.0 12.25B

CerebrumGGUF by mradermacherUpdated 2024-05-06
Download on Hugging Face
Parameters12.5B
Context32K
Layers56
Architecturellama
Smallest quant4.6 GB
Quick answer

Cerebrum 1.0 12.25B needs about 9.9 GB of memory at Q4_K_M (a 7.5 GB file) with an 8K context window, so it fits on a 12 GB GPU. This repository has 14 quants, from 4.6 GB (Q2_K) to 13.3 GB (Q8_0). Longer context needs more: at 32K tokens the KV cache alone adds about 7.5 GB.

Cerebrum 1.0 12.25B – a 12.5B‑parameter LLM for reasoning tasks

Cerebrum 1.0 12.25B is a 12.5B‑parameter model based on the llama architecture. It is a layer‑interleaved merge of the 7B version and is designed for reasoning and chain‑of‑thought tasks. The model is released under the apache‑2.0 license.

Good at

  • Reasoning and chain‑of‑thought tasks
  • Zero‑shot performance on math and logic benchmarks
  • Fine‑tuning foundation for larger models

Limits

  • No vision support
  • Requires 32k context length
  • Large VRAM needed for higher‑quality quants

Best for

  • Academic research requiring reasoning
  • Fine‑tuning for specialized domains
  • Educational tutoring systems

Who it’s for. Researchers and developers who need a large language model focused on reasoning and chain‑of‑thought capabilities. The model is suitable for those who plan to fine‑tune or use it as a drop‑in upgrade from the 7B version.

Quants & memory

QuantFileEst. memoryvs VRAMResultDownload
Q2_K4.6 GB 7.0 GB
Fits Get file →
IQ3_XS5.2 GB 7.5 GB
Tight Get file →
Q3_K_S5.4 GB 7.8 GB
Tight Get file →
IQ3_S5.4 GB 7.8 GB
Tight Get file →
IQ3_M5.6 GB 8.0 GB
Too big Get file →
Q3_K_M6.0 GB 8.4 GB
Too big Get file →
Q3_K_L6.6 GB 8.9 GB
Too big Get file →
IQ4_XS6.8 GB 9.1 GB
Too big Get file →
Q4_K_S7.1 GB 9.5 GB
Too big Get file →
Q4_K_M7.5 GB 9.9 GB
Too big Get file →
Q5_K_S8.6 GB 11.0 GB
Too big Get file →
Q5_K_M8.8 GB 11.2 GB
Too big Get file →
Q6_K10.2 GB 12.6 GB
Too big Get file →
Q8_013.3 GB 15.6 GB
Too big Get file →

For 8 GB VRAM, use Q4_K_M (7.51 GB file, ~9.9 GB memory) or Q3_K_M (6.03 GB file, ~8.4 GB memory). For 12 GB VRAM, Q5_K_S (8.6 GB file, ~11 GB memory) or Q5_K_M (8.83 GB file, ~11.2 GB memory) are appropriate. For 16 GB VRAM, Q6_K (10.24 GB file, ~12.6 GB memory) or Q8_0 (13.26 GB file, ~15.6 GB memory) can be used.

What GPU do you need for Cerebrum 1.0 12.25B?

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: the best quant that fits is IQ3_M (5.6 GB file, about 8.0 GB in use).
  • 12 GB VRAM: the best quant that fits is Q5_K_M (8.8 GB file, about 11.2 GB in use).
  • 16 GB VRAM: the best quant that fits is Q8_0 (13.3 GB file, about 15.6 GB in use).

Which Cerebrum 1.0 12.25B quant should you choose?

Q4_K_M is the usual starting point for Cerebrum 1.0 12.25B: it balances file size and quality well. Here it is a 7.5 GB file.

The Q8_0 file is 13.3 GB, about 1.8 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 4.6 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
Parameters12.5B
Layers56
Trained context32,768 tokens
KV cache (fp16)224 KB per token

Source

Quantized bymradermacher
FamilyCerebrum
Licenseapache-2.0
Downloads (HF)1,539
Last updated2024-05-06

Frequently asked questions about Cerebrum 1.0 12.25B

How much VRAM does Cerebrum 1.0 12.25B need?

About 9.9 GB at Q4_K_M with an 8K context window. That is the 7.5 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 Cerebrum 1.0 12.25B on an 8 GB GPU?

Yes, with IQ3_M: the 5.6 GB file needs about 8.0 GB at 8K context. Larger quants in this repository do not fit in 8 GB.

What is the file size of Cerebrum 1.0 12.25B GGUF?

It depends on the quant: Q2_K is 4.6 GB, IQ3_XS is 5.2 GB, Q3_K_S is 5.4 GB, IQ3_S is 5.4 GB, IQ3_M is 5.6 GB, Q3_K_M is 6.0 GB and 8 more.

What context length does Cerebrum 1.0 12.25B support?

The GGUF header lists a trained context length of 32,768 tokens (32K). The KV cache grows with the context you use: about 1.9 GB at 8K and 7.5 GB at 32K.

Where can I download Cerebrum 1.0 12.25B GGUF?

The files are on Hugging Face in the mradermacher/Cerebrum-1.0-12.25B-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 Cerebrum 1.0 12.25B 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 Cerebrum 1.0 12.25B on 8 GB VRAM?

Yes, with a Q4_K_M or Q3_K_M quant, which require about 9.9 or 8.4 GB of VRAM at 8k context.

Does this model support vision tasks?

No, the model has no vision capabilities.

What context length does it support?

It supports a trained context of 32768 tokens.

Similar models

Browse all →

More from mradermacher

All by mradermacher →

More Cerebrum models

All Cerebrum →

Other medium models

All medium models →