Latest news.
Qwen3.5-122B-A10B and Qwen3.6-35B-A3B go live on ARK Cloud
The two largest Qwen vision-language models are online with a 256K-token context, reasoning and tool calling.
ARK API reaches parity with OpenAI and Anthropic
Tool choice, JSON output, stop sequences, request ids and dozens of details now behave like the APIs developers know.
ARK 2.0 released
A new runtime generation: faster single streams, 64 concurrent streams per model in our tests, and far less energy per token.
Logo and signet
Use the wordmark on light backgrounds and the white version on dark ones. Please do not recolour or distort them.
ARK Labs builds inference infrastructure for Europe. Its platform serves open-weight AI models behind OpenAI- and Anthropic-compatible APIs, on ARK Cloud in the European Union or on the customer's own GPUs with ARK Core and ARK Tailored. ARK Labs is headquartered in Szczecin, Poland, and is backed by Tar Heel Capital and Navivo VC.