ANVE: the private AI office, on the record. See what that means

ANVE

Run it on your own hardware · updated 2026-10-16

bge-m3
The memory it needs, what you may do with it, and the hardware it suits.

Short answer. bge-m3 needs about 1.8 GB of memory at 4-bit, 2.1 GB at 8-bit and 2.6 GB at full 16-bit precision, including runtime overhead. It has about 568 million parameters and an 8,194-token context window. Licence in short: commercial use allowed. By memory alone, on ANVE’s launch configurations: ANVE One fits at BF16, ANVE Office fits at BF16, ANVE Office Pro (GPU) fits at BF16. These are estimates from its published size, not benchmarks.

01

What it is

bge-m3 is an embedding model that turns text into vectors for document search, published by the Beijing Academy of Artificial Intelligence (BAAI) on Hugging Face on 27 January 2024. It has about 568 million parameters and a configured context window of 8,194 tokens.

Its card lists the licence as MIT. In short, the licence allows commercial use, modification and redistribution as long as the copyright notice is kept.

Its card does not state a parameter count; about 568 million is inferred from its 2.27 GB 32-bit weights file (pytorch_model.bin, 4 bytes per parameter).

Hugging Face counted about 33,835,381 downloads of it in the 30 days before we checked (7 Oct 2026), and 3,829 people have liked it there.

02

How much memory it needs

The arithmetic: parameters × bits per weight ÷ 8 gives the weights, plus about 1.5 GB for the runtime. These are planning estimates, not measurements.

Inputs are short, so working memory is small next to the weights.

PrecisionWeightsTotal, estimated
BF16 (full precision)1.1 GB2.6 GB
8-bit (Q8_0)0.60 GB2.1 GB
4-bit (Q4_K)0.32 GB1.8 GB
03

Architecture at a glance

SettingValue in config.json
Architecture classXLMRobertaModel
Model typexlm-roberta
Layers24
Hidden size1,024
Attention heads16
Head size64
Vocabulary250,002
Configured context8,194
Stored precisionfloat32
04

Will it fit on ANVE?

These verdicts are arithmetic, not tests, against the specification at launch: ANVE One can give a model about three quarters of its 64 GB of unified memory, while ANVE Office and ANVE Office Pro hold it on a 24 GB graphics card and keep the rest of the stack in 64 GB or 128 GB of system memory. ANVE’s tested default is Qwen 3.8 27B; this model is not yet validated on ANVE.

Configuration (at launch)Memory for the modelVerdict
ANVE Oneabout 48 GB of 64 GB unified memoryfits at BF16
ANVE Office24 GB on the RTX PRO 4000 Blackwell SFF Editionfits at BF16
ANVE Office Pro (GPU)24 GB on the RTX PRO 4000 Blackwellfits at BF16
05

What an office would use it for

For an office, embeddings are what make "find the clause about late fees" work across thousands of documents. The model is small, so it runs beside the main language model without crowding it.

06

Similar models to compare

  • Qwen3-Embedding-0.6B (0.6B parameters, about 1.8 GB at 4-bit before context; commercial use allowed)
  • embeddinggemma-300m (0.3B parameters, about 1.7 GB at 4-bit before context; commercial use allowed, with conditions)
  • Qwen3-VL-Embedding-2B (2.1B parameters, about 2.7 GB at 4-bit before context; commercial use allowed)
07

Questions people ask

How much memory does bge-m3 need?
About 1.8 GB at 4-bit, 2.1 GB at 8-bit and 2.6 GB at full 16-bit precision, including about 1.5 GB of runtime overhead. These are estimates from its parameter count.
Can bge-m3 run on a 24 GB graphics card?
Yes, at BF16 by this estimate, with room left for context.
Is bge-m3 free for commercial use?
Its card lists MIT, which allows commercial use, modification and redistribution as long as the copyright notice is kept. Read the licence itself before relying on this summary; it is not legal advice.
What is bge-m3’s context window?
Its configuration sets 8,194 tokens. Longer contexts need more memory for the KV cache, so the practical limit on a given machine can be lower.
08

What is not real yet

No sealed ANVE unit exists; every image of the appliance on this site is a labelled render. The software runs today on a development machine. No customers yet, no certifications, no delivery date, and the final price is not announced.

Reservations are ₹999 in India or US$12 elsewhere, refundable in full until you confirm an order in writing (terms). Offices that cannot use cloud AI can apply to be one of ten design partners.