The AI Bench
Various · Differs by model

Embedding models

Small models that turn text into numbers for search and “chat with your documents”. Most run on any computer.

How much memory it needs

Memory to run each model, from its model card or the size of its files. Smaller versions trade some quality for less memory.

ModelVersionNeedsBusiness use
all-MiniLM-L6-v28-bit0.02 GBAllowed
Ollama build0.05 GB
FP320.09 GB
nomic-embed-text v1.5FP160.3 GBAllowed
FP320.6 GB
EmbeddingGemma 300MQuantised0.2 GBAllowed
Ollama build0.6 GB
FP321.2 GB
BGE-M3FP161.2 GBAllowed
FP322.3 GB
Qwen3-Embedding 0.6B8-bit GGUF0.6 GBAllowed
BF161.2 GB
Harrier-OSS-v1 0.6BBF161.2 GBAllowed
EmbeddingGemma 2NVFP40.4 GBAllowed
Ollama build1.3 GB
BF161.5 GB
Qwen3-Embedding 4B4-bit GGUF2.5 GBAllowed
BF168.0 GB
Qwen3-Embedding 8B4-bit GGUF4.7 GBAllowed
BF1615.1 GB

Which computers run it

The best model in this group for each machine.

ComputerBest versionFit
Laptop or PC, no graphics card, 32 GBQwen3-Embedding 8B BF16Runs great
RTX 4060 · 8 GBQwen3-Embedding 8B 4-bit GGUFRuns great
RTX 3060 · 12 GBQwen3-Embedding 8B 4-bit GGUFRuns great
RTX 5060 Ti · 16 GBQwen3-Embedding 8B 4-bit GGUFRuns great
RTX 4090 · 24 GBQwen3-Embedding 8B BF16Runs great
RTX 5090 · 32 GBQwen3-Embedding 8B BF16Runs great
Mac with M4 · 16 GBQwen3-Embedding 8B 4-bit GGUFRuns great
Mac with M4 Pro · 48 GBQwen3-Embedding 8B BF16Runs great
Mac with M5 Max · 128 GBQwen3-Embedding 8B BF16Runs great
Ryzen AI Max+ 395 · 128 GBQwen3-Embedding 8B BF16Runs great
DGX Spark · 128 GBQwen3-Embedding 8B BF16Runs great
RTX PRO 6000 · 96 GBQwen3-Embedding 8B BF16Runs great
Mac Studio with M5 Ultra · 256 GBQwen3-Embedding 8B BF16Runs great
Mac Studio with M3 Ultra · 512 GBQwen3-Embedding 8B BF16Runs great

What it is good and bad at

Good at

  • Search and document Q&A (RAG)
  • Tiny memory needs

Not good at

  • They do not chat — pair them with a chat model

Start it

  • all-MiniLM-L6-v2: runs in sentence-transformers (also ONNX Runtime, OpenVINO, Ollama (GGUF), Transformers.js). Model card
    ollama pull all-minilm:l6-v2
  • nomic-embed-text v1.5: runs in sentence-transformers (also Ollama (GGUF), ONNX, llama.cpp (nomic-ai GGUF)). Model card
    ollama pull nomic-embed-text:v1.5
  • EmbeddingGemma 300M: runs in sentence-transformers (also Ollama, llama.cpp (ggml-org GGUF), ONNX (onnx-community)). Model card
    ollama pull embeddinggemma:300m
  • BGE-M3: runs in sentence-transformers (also FlagEmbedding, Ollama (GGUF), ONNX). Model card
    ollama pull bge-m3:latest
  • Qwen3-Embedding 0.6B: runs in sentence-transformers (also Transformers, vLLM, Ollama (GGUF)). Model card
    ollama pull qwen3-embedding:0.6b
  • Harrier-OSS-v1 0.6B: runs in sentence-transformers. Model card
  • EmbeddingGemma 2: runs in sentence-transformers (also Transformers, Ollama, llama.cpp (ggml-org GGUF), ONNX (onnx-community)). Model card
    ollama pull embeddinggemma-2:latest
  • Qwen3-Embedding 4B: runs in sentence-transformers (also Transformers, vLLM, Ollama (GGUF)). Model card
    ollama pull qwen3-embedding:4b
  • Qwen3-Embedding 8B: runs in sentence-transformers (also Transformers, vLLM, Ollama (GGUF)). Model card
    ollama pull qwen3-embedding:8b