AS

AstraLM 16B MoE

AstralmGGUF by mradermacherUpdated 2024-05-06
Download on Hugging Face
Parameters15.9B
Context32K
Layers32
Architecturellama
Smallest quant6.0 GB
Quick answer

AstraLM 16B MoE needs about 11.3 GB of memory at Q4_K_M (a 9.8 GB file) with an 8K context window, so it fits on a 12 GB GPU. This repository has 14 quants, from 6.0 GB (Q2_K) to 17.0 GB (Q8_0). Longer context needs more: at 32K tokens the KV cache alone adds about 4.3 GB.

AstraLM 16B MoE – a 15.9B parameter mixture‑of‑experts model

AstraLM 16B MoE is a 15.9B parameter llama‑based model with a 32,768 token context window. It is released under a cc‑by‑nc‑4.0 license and supports coding tasks. The model is available in multiple GGUF quantizations.

Good at

  • Large context handling
  • Coding assistance
  • Mixture‑of‑experts inference

Limits

  • No vision support
  • Requires 8,000 token context for memory estimates
  • CC‑BY‑NC license

Best for

  • Code generation
  • Long‑form text analysis
  • Developer tooling

Who it’s for. Engineers and researchers who need a large‑context LLM for coding and text tasks. The model is suitable for those who can comply with a non‑commercial license.

Quants & memory

QuantFileEst. memoryvs VRAMResultDownload
Q2_K6.0 GB 7.6 GB
Tight Get file →
IQ3_XS6.7 GB 8.2 GB
Too big Get file →
Q3_K_S7.0 GB 8.6 GB
Too big Get file →
IQ3_S7.1 GB 8.6 GB
Too big Get file →
IQ3_M7.2 GB 8.8 GB
Too big Get file →
Q3_K_M7.8 GB 9.4 GB
Too big Get file →
Q3_K_L8.4 GB 10.0 GB
Too big Get file →
IQ4_XS8.8 GB 10.3 GB
Too big Get file →
Q4_K_S9.2 GB 10.8 GB
Too big Get file →
Q4_K_M9.8 GB 11.3 GB
Too big Get file →
Q5_K_S11.1 GB 12.7 GB
Too big Get file →
Q5_K_M11.4 GB 13.0 GB
Too big Get file →
Q6_K13.2 GB 14.8 GB
Too big Get file →
Q8_017.0 GB 18.6 GB
Too big Get file →

For 8 GB VRAM, use Q2_K (7.6 GB) or IQ3_XS (8.2 GB). If 12 GB VRAM is available, Q4_K_M (11.3 GB) offers a good balance of size and quality. Larger models like Q5_K_M or Q6_K require more memory and are best for high‑end GPUs.

What GPU do you need for AstraLM 16B MoE?

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 Q2_K (6.0 GB file, about 7.6 GB in use).
  • 12 GB VRAM: the best quant that fits is Q4_K_M (9.8 GB file, about 11.3 GB in use).
  • 16 GB VRAM: the best quant that fits is Q6_K (13.2 GB file, about 14.8 GB in use).
  • 24 GB VRAM: the best quant that fits is Q8_0 (17.0 GB file, about 18.6 GB in use).

Which AstraLM 16B MoE quant should you choose?

Q4_K_M is the usual starting point for AstraLM 16B MoE: it balances file size and quality well. Here it is a 9.8 GB file.

The Q8_0 file is 17.0 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 6.0 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
Parameters15.9B (approximate, mixture of experts)
Layers32
Trained context32,768 tokens
KV cache (fp16)128 KB per token

Source

Quantized bymradermacher
FamilyAstralm
Downloads (HF)614
Last updated2024-05-06

Frequently asked questions about AstraLM 16B MoE

How much VRAM does AstraLM 16B MoE need?

About 11.3 GB at Q4_K_M with an 8K context window. That is the 9.8 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 AstraLM 16B MoE on an 8 GB GPU?

Yes, with Q2_K: the 6.0 GB file needs about 7.6 GB at 8K context. Larger quants in this repository do not fit in 8 GB.

What is the file size of AstraLM 16B MoE GGUF?

It depends on the quant: Q2_K is 6.0 GB, IQ3_XS is 6.7 GB, Q3_K_S is 7.0 GB, IQ3_S is 7.1 GB, IQ3_M is 7.2 GB, Q3_K_M is 7.8 GB and 8 more.

What context length does AstraLM 16B MoE support?

The GGUF header lists a trained context length of 32,768 tokens (32K). The KV cache grows with the context you use: about 1.1 GB at 8K and 4.3 GB at 32K.

Where can I download AstraLM 16B MoE GGUF?

The files are on Hugging Face in the mradermacher/AstraLM-16B-MoE-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 AstraLM 16B MoE use?

Hugging Face lists the license as cc-by-nc-4.0. 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 Q2_K on 8 GB VRAM?

Yes, Q2_K requires approximately 7.6 GB of VRAM at an 8,000 token context.

Is the model licensed for commercial use?

No, the model is licensed under cc‑by‑nc‑4.0, which allows non‑commercial use only.

Does the model support image input?

No, the model does not have vision capabilities.

Similar models

Browse all →

More from mradermacher

All by mradermacher →

Other medium models

All medium models →