Nemotron Nano 12B VL API

NVIDIA, open modelnemotron-nano-12b-v2-vlModel source: Hugging Face

Reads scanned documents and photos: invoices, contracts, forms.

Nemotron Nano 12B VL is NVIDIA's small vision-language model: it reads photos, screenshots and scanned documents and returns the text and the data inside them. On GPU Flow it costs €0.06 per million input tokens and €0.24 per million output.

Reads in one go

≈ 280 pages

131,072 tokens

Brain size

Small and fast

12B VL, NVFP4

Where it runs

Spain (EU)

MIG B200

Understands

Text and images

Vision, OCR

What it is good at

OCR on invoices, contracts, forms and receipts. Turns scanned paperwork into text you can search and process.
Extracting fields into JSON. Pulls out the totals, dates and names you need, ready to go into your system.
Reading screenshots and photos. Understands what is on screen or in the picture, not just the text.
Long documents. A context of 131,072 tokens, about 280 pages.

For awkward layouts (tables, columns, handwriting), step up to Qwen3-VL 32B OCR.

Pricing

€ per million tokens, VAT included

You send (input)0.06 €

It answers (output)0.24 €

1,000 chat answers~0.10 €

Analysing a 200 page contract~0.01 €

try it with 5 € freeActivated with your card. Enough for about 49,020 answers.

For your technical team

Compatible with the OpenAI SDK: change the base URL and the key, and that is it.

curl https://api.gpuflow.ai/v1/chat/completions \
  -H "Authorization: Bearer $GPUFLOW_API_KEY" \
  -d '{ "model": "nemotron-nano-12b-v2-vl",
        "messages": [{ "role": "user", "content": "Hola" }] }'

Context and limits

Context: 131,072 tokens per request, about 280 pages.
Requests per minute: 1,000 per key, with bursts up to double; raisable to 5,000 from the dashboard. 5,000 in total per account.
Tokens per minute: 2,000,000 input (cached tokens do not count) and 400,000 output, per key.
Prompt per request: Up to 256,000 tokens, raisable on request.
If you go over: A 429 error with a Retry-After header telling you how many seconds to wait.

Nemotron Nano 12B VL compared with similar models

	Nemotron Nano 12B VL	Qwen3-VL 32B OCR	Gemma 4 26B A4B
Best for	Reading scans and images at volume	Documents with tables, columns and handwriting	Text and images, reasoning and tools at low cost
Context	≈ 280 pages	≈ 140 pages	≈ 70 pages
Input / output, € per million	0.06 / 0.24	0.10 / 0.35	0.12 / 0.30
1,000 chat answers	~0.10 €	~0.16 €	~0.15 €

Nemotron Nano 12B VL for cost at volume, Qwen3-VL 32B OCR for difficult layouts, Gemma 4 26B when you need text reasoning on top of the image.

Frequently asked questions

Is Nemotron Nano 12B VL free?

The model weights are open source; running them costs GPU time. On GPU Flow you pay €0.06 per million input tokens, and every new account gets €5 of credit when it adds a card.

What does the Nemotron Nano 12B VL API cost?

€0.06 per million input tokens and €0.24 per million output. VAT included, no monthly fee.

Can I use Nemotron Nano 12B VL with Claude Code or Cursor?

Yes. Claude Code connects through the Anthropic-compatible endpoint (/v1/messages); Cursor and other tools through the OpenAI-compatible one (/v1/chat/completions).

Where does it run?

On our NVIDIA B200 GPUs, in a data centre in Spain.

How are images billed?

Each image counts as input tokens, at the same per-million price as text. There is no separate per-image fee.

How does it handle handwritten text?

It is built for clean scans at high volume. For handwriting, tables or multi-column pages, Qwen3-VL 32B OCR is the better fit.