snowflake arctic embed m v1.5 needs about 0.9 GB of memory at Q8_0 (a 0.1 GB file) with an 8K context window, so it fits on a 4 GB GPU. This repository has 6 quants, from 0.1 GB (TQ1_0) to 0.4 GB (F32).
Snowflake Arctic-Embed-M v1.5 – a compact, high‑quality text embedding model
Snowflake Arctic-Embed-M v1.5 is a 109‑million‑parameter BERT‑based model focused on producing highly compressible embedding vectors. It improves upon the original Arctic-Embed-M with better performance and retains retrieval quality even at 128‑byte vectors.
Good at
- High‑quality embeddings
- Strong retrieval on MTEB Retrieval
- Efficient compression to 128 bytes
Limits
- 512‑token context window
- No vision or generation capabilities
- Apache‑2.0 license
Best for
- Text similarity search
- Semantic clustering
- Embedding‑based retrieval systems
Who it’s for. Developers needing compact embeddings for large‑scale search or clustering. Suitable for edge deployments where storage and memory are limited.
Quants & memory
| Quant | File | Est. memory | vs VRAM | Result | |
|---|---|---|---|---|---|
| TQ1_0 | 0.1 GB | 0.6 GB | Fits | Get file → | |
| TQ2_0 | 0.1 GB | 0.6 GB | Fits | Get file → | |
| Q8_0 | 0.1 GB | 0.6 GB | Fits | Get file → | |
| BF16 | 0.2 GB | 0.7 GB | Fits | Get file → | |
| F16 | 0.2 GB | 0.7 GB | Fits | Get file → | |
| F32 | 0.4 GB | 1.0 GB | Fits | Get file → |
For 8 GB VRAM, use TQ1_0 or TQ2_0 (≈0.07 GB file, 0.9 GB memory). For 16 GB VRAM, Q8_0 or BF16 (≈0.12–0.22 GB file, 0.9–1 GB memory). Higher precision F16/F32 offers marginal quality gains at the cost of 1–1.2 GB memory.
What GPU do you need for snowflake arctic embed m v1.5?
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: the best quant that fits is F32 (0.4 GB file, about 1.2 GB in use).
Which snowflake arctic embed m v1.5 quant should you choose?
Q8_0 (0.1 GB) is the middle option in this repository. Close to lossless. It costs roughly twice the memory of a 4-bit file, so it suits cases where you have plenty of VRAM.
If memory is tight, the smallest file is TQ1_0 at 0.1 GB, with the quality trade-offs that come with fewer bits. Lower bit counts mean smaller files and more quality loss.
New to quant names? Read GGUF quantization explained.
Specifications
Architecture from the GGUF header
Source
Frequently asked questions about snowflake arctic embed m v1.5
How much VRAM does snowflake arctic embed m v1.5 need?
About 0.9 GB at Q8_0 with an 8K context window. That is the 0.1 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 snowflake arctic embed m v1.5 on an 8 GB GPU?
Yes, with F32: the 0.4 GB file needs about 1.2 GB at 8K context. Larger quants in this repository do not fit in 8 GB.
What is the file size of snowflake arctic embed m v1.5 GGUF?
It depends on the quant: TQ1_0 is 0.1 GB, TQ2_0 is 0.1 GB, Q8_0 is 0.1 GB, BF16 is 0.2 GB, F16 is 0.2 GB, F32 is 0.4 GB.
What context length does snowflake arctic embed m v1.5 support?
The GGUF header lists a trained context length of 512 tokens (512). The KV cache grows with the context you use: about 0.3 GB at 8K and 0.0 GB at 512.
Where can I download snowflake arctic embed m v1.5 GGUF?
The files are on Hugging Face in the Snowflake/snowflake-arctic-embed-m-v1.5 repository. The download buttons on this page open the file on Hugging Face. ggufmodels.com does not host any model files.
What license does snowflake arctic embed m v1.5 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 this model on 8 GB VRAM?
Yes, the TQ1_0 and TQ2_0 quantizations require about 0.9 GB of VRAM at 8‑k context.
Does this model support image inputs?
No, it is a text‑only embedding model with no vision capabilities.
What compression levels are recommended?
Uniform scalar quantization ranges of –0.18 to +0.18 for 4‑bit and –0.3 to +0.3 for 8‑bit give good performance down to 128‑byte vectors.