LOADING DATASHEET
LOADING DATASHEET
by Baai
RANKED #148 OF 228 ASSISTANTS · OVERALL #238 OF 6,618 · VIBE SCORE 5.7 · 37 VOICES
Reranking model for improving retrieval quality in search and recommendation systems
5 mentions
1 mentions
26 weeks · 25 voices
HONEYMOON FADING
Early praise is cooling off in recent weeks.
Open weights. You can use a hosted service, or download it and run it yourself, free.
Open weights: download Bge Reranker V2 M3 once and it's yours. No subscription, no rate limits, works offline.
568M parameters, FP16 weights take about 1.1 GB (quantized INT8/ONNX takes ~570 MB)
Intel Core i5-12400 / 16 GB RAM PC
16 GB SYSTEM RAM
~$350 USEDCPU inference via PyTorch / FlagEmbedding or ONNX INT8, ~85 tok/s, 8k context
NVIDIA GeForce RTX 3060 12GB
12 GB VRAM
~$260 USEDFP16 full precision via Text Embeddings Inference (TEI) or FlagEmbedding, ~2200 tok/s, 8k context
NVIDIA GeForce RTX 4070 12GB
12 GB VRAM
~$520 NEWFP16 high-throughput batch inference via TEI or vLLM, ~6500 tok/s, 8k context
The three cards above are researched picks. Search the machine you actually have — we only say yes if it should reply at a usable speed, not just load the weights and crawl.
HARDWARE PICKS & PRICES ARE RESEARCHED FROM THE LIVE WEB AND REFRESHED AUTOMATICALLY. TREAT THEM AS BALLPARK, NOT GOSPEL.
Everyday jobs on each of those rigs: how long Bge Reranker V2 M3 takes, and how good the crowd says it is at that kind of work.
| THE JOB | BARE MINIMUMIntel Core i5-12400 / 16 GB RAM PC | SWEET SPOTNVIDIA GeForce RTX 3060 12GB | FULL POWERNVIDIA GeForce RTX 4070 12GB |
|---|---|---|---|
| Write an emaila paragraph or two, drafted from a one-line brief | ~6 S | UNDER A SECOND | UNDER A SECOND |
| Summarize a documenta long report boiled down to the points that matter | ~18 S★★★★★ | UNDER A SECOND★★★★★ | UNDER A SECOND★★★★★ |
| Build a websitea small landing page, markup and styles together | ~1.6 MIN★★★★★ | ~4 S★★★★★ | ~1 S★★★★★ |
| Build a backendan API with routes, storage and tests | ~4.9 MIN★★★★★ | ~11 S★★★★★ | ~4 S★★★★★ |
| Build a gamea playable browser game in one file | ~2.9 MIN★★★★★ | ~7 S★★★★★ | ~2 S★★★★★ |
STARS ARE THE COMMUNITY SCORE FOR THAT KIND OF WORK, DOCKED FOR HOW SQUASHED THE WEIGHTS GET AT EACH TIER. TIMES ARE COMPUTED FROM RESEARCHED SPEEDS. BALLPARK, NOT GOSPEL.
Not enough discussion about Bge Reranker V2 M3 yet to call it. We found 37 posts but people did not say much either way.
28 POSTS · 9 COMMENTS · REDDIT · GITHUB
Other models the crowd has fully reviewed, starting with text models like this one.
I Killed RAG Hallucinations Almost Completely
Is Adding a Reranker to My RAG Stack Actually Worth the Extra Latency? (Explained Simply)
How I Self-Hosted a Local Reranker for Open WebUI with vLLM (No More Jina API)
Best RAG choice on Open-WEBUI
which model to use on local 24gb mac mini M4 pro
We need to talk about Ollama’s lack of reranker support.
Help for RAG
Need Practical Information About SOTA Effective RAG
New external reranking feature in 0.6.9 doesn’t seem to function at all (verified by using Ollama PS)
Best Local RAG Setup for Internal PDFs? (RTX 6000 24GB 256GB RAM i9-10980XE)
How to do basic Documents configuration in OpenWebUI? Going from older version to new version and things have changed.
What we learned deploying RAG for regulated industries (manufacturing, legal, healthcare)