Microsoft Deploys In‑House MAI Models to Lower AI Costs, While Qonto Launches AI‑Powered Finance Integration

Microsoft has begun replacing external AI services from OpenAI and Anthropic with its internally‑developed MAI family of models. The move targets tens of thousands of weekly tasks in Excel and Outlook, aiming to curb the billions of tokens and associated fees paid to third‑party providers. At the Build 2026 conference the company unveiled seven MAI variants, including MAI‑Thinking‑1 (35 billion parameters, 256,000‑token context) and MAI‑Code‑1‑Flash, which it says matches or exceeds the performance of leading rivals such as Claude Opus 4.6 and GPT‑5.5 while delivering a fraction of the cost.

French fintech Qonto introduced Qonto MCP, an integration that connects its banking platform to AI assistants such as Claude and ChatGPT via the Model Context Protocol (MCP). Users can query balances, list recent transactions, create invoices and manage cards using natural‑language prompts, with role‑based permissions and full audit trails. The feature built on early community adoption—380 unofficial users before launch—illustrating a shift toward conversational interfaces that embed financial management directly in AI assistants.