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

ANVE

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

granite-4.0-350m
The memory it needs, what you may do with it, and the hardware it suits.

Short answer. granite-4.0-350m needs about 1.9 GB of memory at 4-bit, 2.1 GB at 8-bit and 2.4 GB at full 16-bit precision, including runtime overhead and an 8,192-token KV cache. It has about 352 million parameters and a 32,768-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

granite-4.0-350m is a language model for chat, drafting and question answering, published by IBM on Hugging Face on 7 October 2025. It has about 352 million parameters and a configured context window of 32,768 tokens.

Its card lists the licence as Apache 2.0. In short, the licence allows commercial use, modification and redistribution, with attribution and a copy of the licence, and includes a patent grant.

Hugging Face counted about 24,170 downloads of it in the 30 days before we checked (7 Oct 2026), and 71 people have liked it there.

It is built on ibm-granite/granite-4.0-350m-base, according to its card.

02

How much memory it needs

How these figures are worked out: weights = parameters × bits per weight ÷ 8. The KV cache is added for 8,192 tokens of context, plus about 1.5 GB of runtime overhead. Treat the totals as planning estimates.

Each token of context adds about 0.03 MB of KV cache at 16-bit, so 32,768 tokens would add about 0.94 GB.

PrecisionWeightsKV cache (8,192 tokens)Total, estimated
BF16 (full precision)0.70 GB0.23 GB2.4 GB
8-bit (Q8_0)0.37 GB0.23 GB2.1 GB
4-bit (Q4_K)0.20 GB0.23 GB1.9 GB
03

Architecture at a glance

SettingValue in config.json
Architecture classGraniteMoeHybridForCausalLM
Model typegranitemoehybrid
Layers28
Hidden size1,024
Attention heads16
Key-value heads4 (grouped-query attention)
Head size64
Vocabulary100,352
Configured context32,768
Stored precisionbfloat16
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

Models this small are best for routing, classification, autocomplete and simple extraction. They are not reliable for minutes or legal drafting: ANVE never uses anything smaller than a 4B-class model for minutes.

06

Similar models to compare

  • LFM2.5-350M (0.4B parameters, about 1.7 GB at 4-bit before context; custom licence: check before commercial use)
  • LFM2-350M (0.4B parameters, about 1.7 GB at 4-bit before context; custom licence: check before commercial use)
  • SmolLM2-360M-Instruct (0.4B parameters, about 1.7 GB at 4-bit before context; commercial use allowed)
07

Questions people ask

How much memory does granite-4.0-350m need?
About 1.9 GB at 4-bit, 2.1 GB at 8-bit and 2.4 GB at full 16-bit precision, including an 8,192-token KV cache and about 1.5 GB of runtime overhead. These are estimates from its parameter count.
Can granite-4.0-350m run on a 24 GB graphics card?
Yes, at BF16 by this estimate, with room left for context.
Is granite-4.0-350m free for commercial use?
Its card lists Apache 2.0, which allows commercial use, modification and redistribution, with attribution and a copy of the licence, and includes a patent grant. Read the licence itself before relying on this summary; it is not legal advice.
What is granite-4.0-350m’s context window?
Its configuration sets 32,768 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.