Which models your team plans include
One main model for chat, agents and code, plus specialist models for speech, decisions and search. All of them run in the EU, the specialist models even in Germany. Every team member reaches them with their own API key through the same API.
At a glance
| Model | Used for | Officework | Advanced | Coding & Automation |
|---|---|---|---|---|
| DeepSeek V4.1 Flash | Chat, agents, coding, long documents | Included | Included | Included |
| Whisper large-v3 turbo | Speech to text, 99 languages | Not included | Included | Included |
| decider-4b | Classify and decide in automations | Not included | Included | Included |
| EmbeddingGemma 2 | Search across your documents (RAG) | Not included | Not included | Included |
Main model
DeepSeek V4.1 Flash
The model you chat with, code with and run agents on. In team plans with up to 1M tokens of context.
- Officework
- Advanced
- Coding & Automation
- Context
- 500,000 to 1,000,000 tokens, depending on the plan
- Interface
- OpenAI- and Anthropic-compatible
- Clients
- Weldall, Claude Code, opencode and more
- Operation
- In the EU
Everyday examples
Contracts and manuals in one go
Whole documents in a single request, ask questions about them, find deviations.
Coding agents
Read repositories, propose and make changes, write tests.
Everyday office writing
Draft proposals, summarize emails, prepare research.
Transcription
Whisper large-v3 turbo
Turns audio into text. A fast OpenAI model that we operate ourselves.
- Not included in Officework
- Advanced
- Coding & Automation
- Languages
- 99
- Can do
- Transcribe (99 languages)
- Origin
- OpenAI, MIT license
- Operation
- In Germany
Everyday examples
Meeting minutes
Recording in, text out, then have DeepSeek summarize it.
Voice notes
Dictate on the go, keep working with the text at the office.
Interviews and customer calls
Write them up for analysis and quotes.
Decision model
decider-4b
Not a chat model: it receives a situation and a question with fixed answer options and returns a probability for each option, in a single fast step.
- Not included in Officework
- Advanced
- Coding & Automation
- Input
- Text and up to 4 images, questions with options
- Output
- Probability per option
- Origin
- Mapika, based on Qwen3.5, Apache 2.0 license
- Operation
- In Germany
Everyday examples
Route tickets
“Which team is responsible: sales, support or accounting?” with confidence in percent.
Rate urgency
Sort incoming requests into low, medium, high before a person looks at them.
Next step in an agent
Should the agent search, ask or answer? Decided quickly and reproducibly.
Embeddings
EmbeddingGemma 2
Turns content into number vectors so you search by meaning instead of keyword. Text, code, images, audio and video end up in the same space.
- Not included in Officework
- Not included in Advanced
- Coding & Automation
- Dimensions
- 768, can be shortened to 512, 256 or 128
- Input
- up to 8,192 tokens, 100+ languages
- Origin
- Google DeepMind, Apache 2.0 license
- Operation
- In Germany
Everyday examples
Your own knowledge search
Make manuals, wiki and tickets searchable and have DeepSeek answer from them.
Find similar cases
Show matching past solutions for a new ticket.
Spot duplicates
Detect duplicate customer requests or records.
Learn more
Endpoints, parameters and limits are in the API docs. Questions about a model, or want a different one? Talk to us.