Mitii Blog

The official source for updates from the Mitii team. Release notes, benchmark results, deprecation notices, and the design decisions behind them all land here first.

Benchmark Results: Mitii AI Agent with Qwen 3.8 27B (45k Context Window)

Jan 15, 2026

We pushed Mitii through 170 non-trivial, full-stack engineering tasks with a 45,000-token context window. Overall pass rate: 89.41% — a GO signal on the overall gate.

Read more