kalle07
embedder_collection
Embedding (not for chat)
- Max context length
- 512
- Architecture
- bert
About embedder_collection
embedder_collection is a model published in GGUF format. GGUF files are published by kalle07/embedder_collection. It has the bert architecture, a maximum context length of 512 tokens (values written in the GGUF).
There are 5 quantizations, with file sizes from 0.03 GB (Q8_0) to 2.27 GB (F32) (actual sizes on Hugging Face).
File size and estimated memory per quantization
- Up to 8 GB
- Up to 16 GB
- Up to 24 GB
- Up to 48 GB
- Up to 128 GB
- More
File sizes are the actual values on Hugging Face. Memory is estimated for each context length (KV cache f16). The formula is in How memory is calculated.
All memory figures are Estimated (KV cache f16). It should run on a setup with at least this much VRAM (or Mac unified memory). Click a value to see the devices and models that fit that amount (Find by GPU / Mac).
About the "quality guide" (how far a quantization drifts from the original)
For each quantization type, how far the output drifts from the original model, grouped into six levels from values llama.cpp measured on Llama-3-8B.
The value is mean KLD (the difference between the original and quantized model in next-token probabilities); smaller means closer to the original. Quantizations made with an imatrix (importance data) come out slightly smaller. Levels use the value without an imatrix (or with one, for types that only have that).
These values were not measured on this model. Larger models and MoE models tend to drift less with the same type. Distributor-specific mixed types (such as UD-Q4_K_XL) cannot be matched by type name alone, so they show "-".
| Quantization | Quality guide | KLD (no imatrix) | KLD (imatrix) | Size (Llama-3-8B) |
|---|---|---|---|---|
| Q8_0 | Nearly identical | 0.0014 | - | 7.96 GiB |
| Q6_K | Very small | 0.0055 | - | 6.14 GiB |
| Q5_K_M | Very small | 0.0108 | - | 5.33 GiB |
| Q5_K_S | Small | 0.0166 | - | 5.21 GiB |
| Q5_1 | Small | 0.0180 | - | 5.65 GiB |
| Q5_0 | Small | 0.0222 | - | 5.21 GiB |
| Q4_K_M | Small | 0.0313 | 0.0282 | 4.58 GiB |
| Q4_K_S | Medium | 0.0431 | 0.0320 | 4.37 GiB |
| IQ4_NL | Small | - | 0.0357 | 4.35 GiB |
| IQ4_XS | Small | - | 0.0363 | 4.14 GiB |
| Q4_1 | Medium | 0.0717 | - | 4.78 GiB |
| Q4_0 | Medium | 0.0719 | - | 4.34 GiB |
| Q3_K_L | Medium | 0.0872 | 0.0731 | 4.03 GiB |
| Q3_K_M | Medium | 0.1019 | 0.0844 | 3.74 GiB |
| IQ3_M | Medium | - | 0.1025 | 3.53 GiB |
| IQ3_S | Medium | - | 0.1113 | 3.42 GiB |
| IQ3_XS | Large | - | 0.1387 | 3.28 GiB |
| IQ3_XXS | Large | - | 0.1836 | 3.05 GiB |
| Q3_K_S | Large | 0.2282 | 0.1998 | 3.41 GiB |
| IQ2_M | Large | - | 0.3260 | 2.74 GiB |
| Q2_K | Very large | 0.4451 | 0.3322 | 2.96 GiB |
| Q2_K_S | Very large | - | 0.4034 | 2.96 GiB |
| IQ2_S | Very large | - | 0.4392 | 2.56 GiB |
| IQ2_XS | Very large | - | 0.5463 | 2.43 GiB |
| IQ2_XXS | Very large | - | 0.8120 | 2.24 GiB |
| IQ1_M | Very large | - | 1.3931 | 2.01 GiB |
| IQ1_S | Very large | - | 2.2546 | 1.88 GiB |
Source: llama.cpp (MIT), measured on Llama-3-8B against FP16, Wikitext-2, revision f364eb6f. Original table · Reading and choosing quantizations
Last checked: Oct 18, 2026
- Architecture
- bert
- Supported before llama.cpp b4409
- Max context length
- 512
| Quantization | File size Measured | Quality guide | 512 max |
|---|---|---|---|
| Q8_0 minilmv2-toxic-jigsaw-q8_0 | 0.03 GB | Nearly identical | 1.10 GB |
| F16 all-minilm-l6-v2_f16 | 0.05 GB | Original precision | 1.12 GB |
| F32 all-minilm-l6-v2_f32 | 0.09 GB | Original precision | 1.16 GB |
| Q8_0 toxic-prompt-roberta-q8_0 | 0.14 GB | Nearly identical | 1.21 GB |
| Q8_0 jina-embeddings-v2-base-code-Q8_0 | 0.17 GB | Nearly identical | 1.25 GB |
| F16 paraphrase-multilingual-MiniLM-L12-118M-v2-F16 | 0.24 GB | Original precision | 1.32 GB |
| F16 granite-embedding-125m-english-f16 | 0.25 GB | Original precision | 1.32 GB |
| F16 nomic-embed-text-v1.5_f16 | 0.27 GB | Original precision | 1.35 GB |
| F16 jina-embeddings-v2-base-en-f16 | 0.27 GB | Original precision | 1.35 GB |
| Q8_0 cross-en-de-es-roberta-sentence-transformer-q8_0 | 0.30 GB | Nearly identical | 1.38 GB |
| Q8_0 cross-en-de-fr-roberta-sentence-transformer-q8_0 | 0.30 GB | Nearly identical | 1.38 GB |
| Q8_0 bce-embedding-base_v1-q8_0 | 0.30 GB | Nearly identical | 1.38 GB |
| Q8_0 german-roberta-sentence-transformer-v2-q8_0 | 0.30 GB | Nearly identical | 1.38 GB |
| Q8_0 cross-en-de-roberta-sentence-transformer-fix-q8_0 | 0.30 GB | Nearly identical | 1.38 GB |
| Q8_0 paraphrase-multilingual-mpnet-base-v2-q8_0 | 0.30 GB | Nearly identical | 1.38 GB |
| F16 jina-embeddings-v2-base-de-f16 | 0.32 GB | Original precision | 1.40 GB |
| Q8_0 sentence-transformers-e5-large-v2-q8_0 | 0.36 GB | Nearly identical | 1.43 GB |
| Q8_0 mug-b-1.6-q8_0 | 0.36 GB | Nearly identical | 1.43 GB |
| Q8_0 bge-large-en-v15-q8_0 | 0.36 GB | Nearly identical | 1.43 GB |
| Q8_0 mxbai-embed-2d-large-v1-q8_0 | 0.36 GB | Nearly identical | 1.43 GB |
| Q8_0 bge-large-mpnet-base-all-nli-triplet-final-q8_0 | 0.36 GB | Nearly identical | 1.43 GB |
| Q8_0 mxbai-embed-large-v1.Q8_0 | 0.36 GB | Nearly identical | 1.43 GB |
| Q8_0 gist-large-embedding-v0-Q8_0 | 0.36 GB | Nearly identical | 1.43 GB |
| Q8_0 bge-m3_en_ru-q8_0 | 0.41 GB | Nearly identical | 1.48 GB |
| Q4_K_M | 0.41 GB | Small | 1.48 GB |
| Q8_0 nomic-embed-text-v2-moe.Q8_0 | 0.51 GB | Nearly identical | 1.59 GB |
| Q8_0 kalm-embedding-multilingual-mini-instruct-v2-q8_0 | 0.53 GB | Nearly identical | 1.60 GB |
| F32 nomic-embed-text-v1.5.f32 | 0.55 GB | Original precision | 1.62 GB |
| BF16 | 0.56 GB | Original precision | 1.63 GB |
| F16 bce-embedding-base_v1-f16 | 0.56 GB | Original precision | 1.64 GB |
| F16 granite-embedding-278m-multilingual-f16 | 0.56 GB | Original precision | 1.64 GB |
| F16 paraphrase-multilingual-mpnet-base-277M-v2-F16 | 0.56 GB | Original precision | 1.64 GB |
| Q8_0 sentence-transformers-multilingual-e5-large-q8_0 | 0.60 GB | Nearly identical | 1.68 GB |
| Q8_0 multilingual-e5-large-instruct-q8_0 | 0.60 GB | Nearly identical | 1.68 GB |
| Q8_0 multilingual-e5-large-instruct-xnli-q8_0 | 0.60 GB | Nearly identical | 1.68 GB |
| F16 embeddinggemma-300M-F16 | 0.61 GB | Original precision | 1.69 GB |
| Q8_0 bge-m3-q8_0 | 0.63 GB | Nearly identical | 1.71 GB |
| Q8_0 snowflake-arctic-embed-l-v2.0-q8_0 | 0.63 GB | Nearly identical | 1.71 GB |
| Q8_0 ger-RAG-bge-M3-merg-snowf-artic-hessian-AI.Q8_0 | 0.63 GB | Nearly identical | 1.71 GB |
| Q8_0 bge-reranker-v2-m3-q8_0 | 0.64 GB | Nearly identical | 1.71 GB |
| Q8_0 Qwen3-Embedding-0.6B-Q8_0 | 0.64 GB | Nearly identical | 1.71 GB |
| F16 bge-large-en-v1.5-f16 | 0.67 GB | Original precision | 1.74 GB |
| F16 mxbai-embed-large-v1-f16 | 0.67 GB | Original precision | 1.74 GB |
| F16 gte-large.f16 | 0.67 GB | Original precision | 1.74 GB |
| F16 e5-large-v2.f16 | 0.67 GB | Original precision | 1.74 GB |
| F16 gist-large-Embedding-v0.f16 | 0.67 GB | Original precision | 1.74 GB |
| F16 nomic-embed-text-v2-moe.f16 | 0.96 GB | Original precision | 2.03 GB |
| F16 HIT-TMG_KaLM-embedding-multilingual-mini-instruct-v1.5_qwenbased_f16 | 0.99 GB | Original precision | 2.07 GB |
| F32 cross-en-de-fr-roberta-sentence-transformer_F32 | 1.12 GB | Original precision | 2.19 GB |
| F32 cross-en-de-es-roberta-sentence-transformer_F32 | 1.12 GB | Original precision | 2.19 GB |
| F32 german-roberta-sentence-transformer-v2.F32 | 1.12 GB | Original precision | 2.19 GB |
| F32 cross-en-de-roberta-sentence-transformer.F32 | 1.12 GB | Original precision | 2.19 GB |
| F16 jina-embeddings-558M-v3-F16 | 1.12 GB | Original precision | 2.20 GB |
| F16 multilingual-e5-large-instruct-F16 | 1.13 GB | Original precision | 2.20 GB |
| F16 bge-m3.F16 | 1.16 GB | Original precision | 2.23 GB |
| F16 German-RAG-BGE-M3-TRIPLES-HESSIAN-AI.f16 | 1.16 GB | Original precision | 2.23 GB |
| F16 snowflake-arctic-embed-l-v2.0-f16 | 1.16 GB | Original precision | 2.23 GB |
| F16 bge-reranker-v2-m3-FP16 | 1.16 GB | Original precision | 2.23 GB |
| F16 Qwen3-Embedding-0.6B-f16 | 1.20 GB | Original precision | 2.27 GB |
| F16 jina-embeddings-v5-text-small-retrieval-F16 | 1.20 GB | Original precision | 2.27 GB |
| F32 embeddinggemma-300M-F32 | 1.22 GB | Original precision | 2.29 GB |
| F32 gist-large-Embedding-v0.f32 | 1.34 GB | Original precision | 2.41 GB |
| F32 snowflake-arctic-embed-l-v2.0-f32 | 2.27 GB | Original precision | 3.35 GB |
| F32 ger-RAG-bge-M3-merg-snowf-artic-hessian-AI.F32 | 2.27 GB | Original precision | 3.35 GB |
Ollama ollama run hf.co/kalle07/embedder_collection:Q4_K_M
llama.cpp llama-server -hf kalle07/embedder_collection:Q4_K_M
How to run
Example commands using Q4_K_M. To use another quantization, replace the name with one from the table above. How to run local LLMs
Ollama
Run it by pointing directly at the GGUF on Hugging Face.
ollama run hf.co/kalle07/embedder_collection:Q4_K_M
llama.cpp
Pass the repository and quantization name to llama-server (or llama-cli) with -hf.
llama-server -hf kalle07/embedder_collection:Q4_K_M
LM Studio
Paste kalle07/embedder_collection into the search box, download it, and choose the quantization from the file list.
Frequently asked questions
- How much memory do I need to run embedder_collection (Q4_K_M)?
- The file size is 0.41 GB (measured). With a 512-token context, the estimated memory including the KV cache is 1.48 GB. The formula is published at https://localllm.az-arte.com/en/guide/memory-calculation.
- What is the command to run embedder_collection with Ollama?
- Run "ollama run hf.co/kalle07/embedder_collection:Q4_K_M" to use the GGUF on Hugging Face directly. To use another quantization, replace the name at the end.