Decision models · BAAI

BGE Reranker v2 Gemma

Task: Reranking search results
Returns: Probability or score
Precision tier: 78 / 100
Size: 2.5B
Max input: 8K tokens
License: apache-2.0
Released: 2024-03

What it decides

Returns: A relevance score for each query and document pair.
Checks: Query and document pairs
Languages: Multilingual
Precision: 78 / 100 · Gemma 2B-based; stronger than v2 M3 on BEIR in BAAI’s results.
Clarity: Probability or score (15 of 15 clarity points)
How it runs: Scores query and document pairs · a small language model, so it runs in the same apps as LLMs.
License: apache-2.0 (Open (Apache, MIT))

Source: model card on Hugging Face (opens in a new tab) · How decision models are rated

How to run it

Quantized files · GGUF

Works in llama.cpp (llama-server --rerank), Lemonade and other llama.cpp servers with a rerank endpoint. No sign-in needed.

huggingface.co/mradermacher/bge-reranker-v2-gemma-GGUF (opens in a new tab)

llama-server -hf mradermacher/bge-reranker-v2-gemma-GGUF --rerank

Original weights · Hugging Face

Full-precision files from BAAI, for Transformers, Sentence Transformers, Text Embeddings Inference, Infinity or vLLM.

huggingface.co/BAAI/bge-reranker-v2-gemma (opens in a new tab)

No sign-in needed.

Links checked 3 days ago against Hugging Face, Ollama and GitHub.

Apps that run BGE Reranker v2 Gemma

Our pick for each kind of computer, assuming the model fits in its memory. Pick your own GPU or Mac in the Model Finder for an exact answer.

Mac (Apple silicon)

Serves the GGUF build behind a /v1/rerank endpoint with llama-server --rerank.

Also good: Sentence Transformers, Lemonade

Windows PC, NVIDIA GPU

Serves the GGUF build behind a /v1/rerank endpoint with llama-server --rerank.

Also good: Sentence Transformers, Lemonade

Windows PC, AMD Radeon

Serves the GGUF build behind a /v1/rerank endpoint with llama-server --rerank.

Also good: Lemonade, Sentence Transformers

Linux, NVIDIA GPU

Serves the GGUF build behind a /v1/rerank endpoint with llama-server --rerank.

Also good: vLLM, Sentence Transformers

Linux server, 80 GB+ GPUs

Serves the GGUF build behind a /v1/rerank endpoint with llama-server --rerank.

Also good: vLLM, Sentence Transformers

App	Type	Opens it as	Systems	Engine	Start
llama.cpp	CLI + API	GGUF	macOSWindowsLinux	llama.cpp	`llama-server -hf mradermacher/bge-reranker-v2-gemma-GGUF:Q4_K_M --rerank`
vLLM	Server	Original weights	WSLWindowsLinux	vLLM	`vllm serve BAAI/bge-reranker-v2-gemma`
Lemonade	Desktop app	GGUF	macOSWindowsLinux	llama.cpp, ONNX Runtime GenAI	`lemonade pull mradermacher/bge-reranker-v2-gemma-GGUF:Q4_K_M`
Sentence Transformers	Library	Original weights	macOSWindowsLinux	PyTorch	Get Sentence Transformers

Which apps can open a model depends on the files published for it: GGUF for llama.cpp-based apps, MLX for Apple’s engine, the original weights for vLLM and SGLang. Compare every app.

Memory and speed

Each quantization on an NVIDIA GeForce RTX 4090, for one decision on a 512-token input. The best hardware table below covers other GPUs and Macs.

Format	Memory needed	Time per decision	Decisions per second	Quality	Download
`Q4_K_M`	2.2 GB	28 ms	36/s	74Great quality (A) · 74 of 100: how capable the model is at Q4_K_M, whatever it runs on. Match score 88/100 (quality, fit, speed and memory use) sets the ranking.	GGUF (opens in a new tab)
`Q5_K_M`	2.4 GB	28 ms	36/s	75Great quality (A) · 75 of 100: how capable the model is at Q5_K_M, whatever it runs on. Match score 89/100 (quality, fit, speed and memory use) sets the ranking.	GGUF (opens in a new tab)
`Q6_K`	2.7 GB	29 ms	35/s	77Great quality (A) · 77 of 100: how capable the model is at Q6_K, whatever it runs on. Match score 89/100 (quality, fit, speed and memory use) sets the ranking.	GGUF (opens in a new tab)
`Q8_0`	3.3 GB	30 ms	34/s	78Great quality (A) · 78 of 100: how capable the model is at Q8_0, whatever it runs on. Match score 90/100 (quality, fit, speed and memory use) sets the ranking.	GGUF (opens in a new tab)
`F16`	5.7 GB	34 ms	29/s	78Great quality (A) · 78 of 100: how capable the model is at F16, whatever it runs on. Match score 90/100 (quality, fit, speed and memory use) sets the ranking.	GGUF (opens in a new tab) Original weights (opens in a new tab)

One request at a time. Servers that batch (Text Embeddings Inference, vLLM, Infinity) handle many more decisions per second.

Best hardware for BGE Reranker v2 Gemma

The least expensive GPU in our catalog that runs it without offloading is the Intel Arc A380 (launch price $139) at Q8_0.

Quality	Hardware	Quant	Memory	Time per decision	Runs
78Great quality (A) · 78 of 100: how capable the model is at F16, whatever it runs on. Match score 90/100 (quality, fit, speed and memory use) sets the ranking.	AMD Radeon RX 7900 XTX	`F16`	6.9 GB / 24.0 GB	31 ms	Runs great
78Great quality (A) · 78 of 100: how capable the model is at F16, whatever it runs on. Match score 90/100 (quality, fit, speed and memory use) sets the ranking.	AMD Radeon RX 9070 XT	`F16`	6.5 GB / 16.0 GB	41 ms	Runs great
78Great quality (A) · 78 of 100: how capable the model is at F16, whatever it runs on. Match score 90/100 (quality, fit, speed and memory use) sets the ranking.	AMD Ryzen AI Max+ 395 (128 GB)	`F16`	13.3 GB / 96.0 GB	78 ms	Runs great
78Great quality (A) · 78 of 100: how capable the model is at F16, whatever it runs on. Match score 90/100 (quality, fit, speed and memory use) sets the ranking.	Apple M2 Ultra (192 GB)	`F16`	17.2 GB / 144 GB	93 ms	Runs great
78Great quality (A) · 78 of 100: how capable the model is at F16, whatever it runs on. Match score 90/100 (quality, fit, speed and memory use) sets the ranking.	Apple M3 Ultra (256 GB)	`F16`	21.0 GB / 192 GB	89 ms	Runs great
78Great quality (A) · 78 of 100: how capable the model is at F16, whatever it runs on. Match score 90/100 (quality, fit, speed and memory use) sets the ranking.	Apple M3 Ultra (512 GB)	`F16`	36.4 GB / 384 GB	89 ms	Runs great
78Great quality (A) · 78 of 100: how capable the model is at F16, whatever it runs on. Match score 90/100 (quality, fit, speed and memory use) sets the ranking.	NVIDIA GB10 (128 GB)	`F16`	14.9 GB / 115 GB	94 ms	Runs great
78Great quality (A) · 78 of 100: how capable the model is at F16, whatever it runs on. Match score 90/100 (quality, fit, speed and memory use) sets the ranking.	NVIDIA GeForce RTX 3090	`F16`	6.9 GB / 24.0 GB	63 ms	Runs great
78Great quality (A) · 78 of 100: how capable the model is at Q8_0, whatever it runs on. Match score 90/100 (quality, fit, speed and memory use) sets the ranking.	NVIDIA GeForce RTX 4060 Ti 16GB	`Q8_0`	4.1 GB / 16.0 GB	99 ms	Runs great
78Great quality (A) · 78 of 100: how capable the model is at F16, whatever it runs on. Match score 90/100 (quality, fit, speed and memory use) sets the ranking.	NVIDIA GeForce RTX 4070 SUPER	`F16`	6.3 GB / 12.0 GB	72 ms	Runs great

Check it on your own hardware

Estimates come from published hardware specs and our open formulas, not from running the model on your machine. Real speed varies with runtime, drivers and settings. Numbers marked “Measured” come from a linked benchmark.