model catalogue
17 open source models, served in Europe.
Open models, running on our machines in Spain. They connect with the same tools you already use for OpenAI or Anthropic, and your data never leaves the EU. From 0.06 € per million tokens.
NVIDIA B200 GPUs in Spain (EU)
the catalogue at a glance
What is inside, and how to read it
token
The chunk of text models count in. A million tokens is roughly 750,000 words.
reads in one go
How much text it can hold without re-reading it, what we call the context. The bigger it is, the more documents fit at once.
you send, it answers
Input and output. Each has its own price per million tokens, and answering costs more than reading.
the “B” in the name
Brain size, in billions of parameters. More B usually means more capable, and more expensive.
moe
Mixture of Experts: an architecture that splits big tasks into parts and shares them among several specialists, switching on only the ones it needs. That is why a huge one can still be fast and cheap.
text, reasoning and vision
Language models
The ones that read and write text. These are the ones answering in Token Factory.
- DeepSeek V4 Pro
DeepSeek AI
The most capable in the catalogue: reasons through long problems, writes code and drives your tools.
1.6T MoE8× B2001,048,576 contextFP8input and output in € per million
0.80 / 2.60 €View model - GLM 5.3 NVFP4
Z.ai
An all-rounder for writing and coding, with tool use.
753B MoE, 40B active8× B2001,048,576 contextNVFP4input and output in € per million
1.206 / 3.789 €View model - DeepSeek V4 Flash
DeepSeek AI
Fast, with a huge memory: it reads a whole contract without losing the thread.
1.6T MoE2× B200524,288 contextFP8input and output in € per million
0.14 / 0.28 €View model - Qwen3.8 27B
Alibaba Cloud
Strong at coding and at chaining tasks on its own.
27BMIG B20098,304 contextNVFP4input and output in € per million
0.10 / 0.30 €View model - Qwen3 14B
Alibaba Cloud
Thinks before answering, and you decide how much: a balance of quality and spend.
14BMIG B20032,768 contextFP8input and output in € per million
0.20 / 1.00 €View model - Qwen3.5 9B
Alibaba Cloud
Light and very fast, for simple tasks at high volume.
9BMIG B20065,536 contextFP8input and output in € per million
0.10 / 0.15 €View model - ALIA 40B Instruct
Barcelona Supercomputing Center
Spain's public model, trained in Spanish, Catalan, Basque and Galician.
40B denso1× B200163,840 contextBF16input and output in € per million
0.20 / 2.00 €View model - Gemma 4 26B A4B
Google DeepMind
Understands text and images at once without being a heavy model.
26B MoE, 3.8B activeMIG B20032,768 contextNVFP4input and output in € per million
0.12 / 0.30 €View model - Nemotron Nano 12B VL
NVIDIA
Reads scanned documents and photos: invoices, contracts, forms.
12B VLMIG B200131,072 contextNVFP4input and output in € per million
0.06 / 0.24 €View model - Qwen3-VL 32B OCR
Alibaba Cloud
Document reading for awkward layouts: tables, columns and handwriting.
32BMIG B20065,536 contextNVFP4input and output in € per million
0.10 / 0.35 €View model - Nemotron 3.5 Lightning 30B
NVIDIA
Reasons and uses tools while spending very little, built to answer fast.
30B MoE, 3B activeMIG B200131,072 contextNVFP4input and output in € per million
0.12 / 0.30 €View model
voice and audio
Voice: audio into text, and text into voice
Whisper turns audio into text, and synthetic voices (TTS) do the opposite. Billed per use: Whisper per minute of audio, voices per million characters.
- Whisper large-v3
OpenAI
Turns audio into text: meetings, calls, interviews and videos.
1.5BMIG B200price per use
0.0015 € / minView model - Qwen3-TTS CustomVoice
Alibaba Cloud
Reads your text aloud in the voice you choose.
1.7BMIG B200price per use
19.00 € / 1M charsView model - Qwen3-TTS Base (clone)
Alibaba Cloud
Turns text into speech: the cheapest option in the catalogue.
1.7BMIG B200price per use
5.00 € / 1M charsView model - Fish Speech S2 Pro
Fish Audio
Very natural speech, able to imitate the sample you provide.
S2 ProMIG B200price per use
8.00 € / 1M charsView model - Voxtral TTS
Mistral AI
Ready-made Mistral voices to read your text aloud.
4BMIG B200price per use
6.00 € / 1M charsView model
music generation
Music: generate songs from text
ACE-Step turns lyrics and a style into a complete song. It is not instant: you request it, wait, and download it when it is ready. Charged per song.
Trademarks of their respective owners, cited only to identify the models we serve.
Ready to call them?
Create your account with 50 € free. If you code, you change one line of config; if you do not, hand it to whoever handles the technical side.