DONNA family

Cadenza

The powerful general-purpose model of the DONNA family. Designed for speed, reliability and a balanced price-performance ratio in everyday use.

The central tool

Donna Cadenza is the workhorse of the DONNA family — a general-purpose model for the vast majority of tasks. From the daily e-mail through evaluations to structured analysis: Cadenza understands context, thinks in a structured way and delivers precise results — quickly and economically.

For tasks that demand deep, multi-step reasoning, extensive tool orchestration or pronounced multimodality, Sinfonia — the flagship of the family — is available.

Cadenza supports more than 30 languages, with particular attention to official EU and UN languages in our configuration. It not only understands text but can also analyse images, call tools and carry out multi-step reasoning processes.

Cadenza is based on curated, licensed and public data sets — customer data never flows into model adaptation. Inference takes place primarily on European infrastructure (see model specification). Every release undergoes comprehensive safety testing and bias checks.

Use cases

Text generation & summarisation
Data analysis & reporting
Translation & language assistance
Document classification
Research & knowledge extraction
Customer communication
Minutes & transcription
Workflow automation

Limitations & notes

Cadenza can produce factually incorrect, incomplete or inconsistent results. The model has no access to real-time information and is limited to the knowledge state of its training.

Cadenza must not be used as the sole basis for decisions with legal, medical, financial or safety-related consequences. Human review is strongly recommended for all decision-relevant outputs.

Model specification

TypeGPAI
ArchitectureTransformer LLM
ConfigurationPre-/post-conditioned (e.g. adapters, fine-tuning, prompt conditioning)
InputText, vision
OutputText
CapabilitiesReasoning, tool calling
Languages30+ (official EU & UN languages)
InferenceSweden (EU), England (UK, non-EU failover)

Benchmarks

Benchmark results

93%
AIME 2025
Mathematics
88%
GPQA Diamond
Natural sciences (PhD level)
87%
MMLU-Pro
Reasoning & knowledge

Methodology

Internal evaluation of the configuration dbpg DONNA Cadenza, build 20260316 (Q1 2026). As of Q2 2026.

Pricing

Usage & billing

Input
from€4
/ 1M tokens
Output
from€20
/ 1M tokens

Billing is on a prepaid basis in the form of credits. The exchange rate from euros to credits depends on the purchase volume; the prices shown are starting prices. Further information at centra.dbpg.io.

Try Cadenza

Available as a closed preview via the Centra platform.

Request access