Tassili builds compact language models from scratch, domain-specific layers and workflow agents — for enterprises that need automation with control, traceability and real deployment economics.
One family, three sizes. Pick the smallest model that does the job — and keep your cost, latency and data under control.
The smallest model of the family. Built for low-latency, high-volume steps and private execution on modest hardware.
The balanced model for everyday enterprise work: grounded answers, drafting and structured workflow reasoning.
Our largest model, for multi-step reasoning and agentic workflows that combine tools, retrieval and approvals.
Tassili is not a general chatbot company. It is an enterprise model systems company.
We build systems that help enterprises execute work — not just generate text: compact serving, domain control and production-ready agents with clear operational boundaries.
Small language models, trained from scratch and optimised for recurring workflow steps, low latency and cost-effective production.
Domain-specific layers that refine vocabulary, reasoning and behaviour for finance, operations, support and compliance.
Agents that connect models, tools, retrieval and human approvals to complete multi-step business tasks end to end.
Compact models run on less hardware, so every automated step costs less in production.
DSLM layers narrow behaviour to your domain instead of relying on generic prompting.
Scoped agents, explicit permissions and human review on higher-risk actions.
APIs, connectors and private deployment that fit your existing systems.
Run Tassili as a private API, fully on-premise, or in a hybrid setup.
Centralised serving with shared access, policy controls and managed routing.
Local serving for sensitive operations, isolated infrastructure and low latency.
Local SLM execution with routed DSLM or agent services for richer workflows.
See them run on your own workflows. We set up a private demo for your team — on our infrastructure or yours.