Skip to main content
DeepMask gives you more than 50 AI models from 11 providers in a single workspace. You can switch between models at any time without losing your conversation context. Every model runs on EU infrastructure, and many run in Germany for teams with stricter data residency requirements.

Anthropic

Opus 5.5 · Sonnet 5.5 · Haiku 5.5 — The newest Claude generation, with 1M-token context and built-in reasoning.Opus 5 · Sonnet 5 and Opus 4.5–4.8 · Sonnet 4.5 / 4.6 · Haiku 4.5 — Earlier Claude models for teams that rely on them.

OpenAI

GPT-6.1 Sol · GPT-6 Sol · GPT-6 Luna — OpenAI’s newest reasoning models with image input and tool use.GPT-5.4 · GPT-5.2 · GPT-4.1 · GPT-4o · GPT-o3 Mini — Proven models for everyday work, long context, and focused reasoning.GPT-OSS 120B (StackIT / Infercom) — Open-weight model, hosted in Germany or the EU.

Google

Gemini 3.8 · 3.7 · 3.5 Flash and Gemini 3.5 · 3.1 Flash Lite — Fast models with a 1M-token context window.Gemma 4 31B · Gemma 3 27B (StackIT) — Open-weight models hosted in Germany.

Alibaba (Qwen)

Qwen3.8 · Qwen3.8 27B (evroc) · Qwen3.5 397B · Qwen3.5 122B · Qwen3 235B (2507) · Qwen3-VL 30B · Qwen3 (StackIT) — From flagship reasoning to compact vision models, many hosted in Germany.

Moonshot AI (Kimi)

Kimi K3 · Kimi K2.7 Code · Kimi K2.6 (evroc) · Kimi K2.5 · Kimi K2 (DeepMask) — Agent-focused models for long, multi-step tasks, coding, and vision.

Z.ai (GLM)

GLM-5.3 · GLM 5.3 (evroc) · GLM-5 · GLM-4.7 · GLM-4.7 Flash — Text models for reasoning, agentic coding, and fast automation.

DeepSeek

DeepSeek V4 Flash · DeepSeek V3.2 — Cost-efficient text-only reasoning models; V4 Flash has a 1M-token context window.

MiniMax

MiniMax M3 · M2.5 · M2.1 — Mixture-of-Experts models for agents and coding; M3 adds image input and 1M-token context.

Mistral AI

Mistral Large 3 — Open-weight flagship for multilingual reasoning and enterprise workflows.

NVIDIA

Nemotron 3 Super — Efficient text-only reasoning model with tool use.

DeepMask

DeepMask D1 · DeepMask D2 — Routers that pick a suitable model for each message. D2 uses only German-hosted models.
Models marked (StackIT) run on StackIT, the German sovereign cloud operated by the Schwarz Group. Models marked (evroc) run on evroc’s sovereign cloud in Germany. (Infercom) models use EU-hosted endpoints from Infercom, and (DeepMask) models are operated by DeepMask in Germany. If your organization requires German hosting, choose a model marked Germany in the table below, or use DeepMask D2.

All models at a glance

Every model reads documents and, except Gemma 3 27B (StackIT), supports tool use. Context windows and output limits are the limits that apply in DeepMask.
Not sure where to start? See Choosing a model, or let DeepMask D1 pick a model for each message.