Product Update: July 2026
5 MIN. READ
This month, we announced:
Agents: Create agents for specific tasks. Each one has its own instructions, skills, files, and memory, and runs on the best and most cost-efficient model for the job.
Excel Agent Review: An agent reviews your Excel outputs for errors, inconsistencies, broken links, and hardcodes, and flags what needs your eye.
Teams Plug-in: Ask your agent for research, models, or decks without leaving Microsoft Teams.
GPT-5.6: Added to the platform after testing with OpenAI pre-release.
Schedule and Event-Based Agents: Set an agent to run on a schedule, or have it triggered by market events, and have the output waiting in your inbox.
Homepage Redesign: A refreshed homepage that reflects how your workflows have evolved.
LSEG Integration: Access LSEG's trusted, licensed data directly in Model ML through LSEG's MCP connector.
Snowflake Integration: Connect your Model ML agents directly to your Snowflake data.
Model ML MCP: Connect Model ML to Claude, ChatGPT, or any MCP client and let them work against your data.
Multi-file Skills: Teach the agent a whole convention, templates, house style, worked examples, not just a single instruction.
Claude Opus 5: Added to the platform after testing with Anthropic pre-release.
Custom MCP: Fire a workflow the moment a company hits a signal you've configured.
Agents
You can now create agents for specific tasks. Each one has its own instructions, skills, files, memory and model, and appears in the composer alongside your main agent. Create one from scratch, duplicate an existing one, or just describe the job in chat and your main agent builds it.
Excel Agent Review
You can now review your workbooks in Excel with Model ML. Your agent checks for logic errors across tabs, broken references, inconsistent assumptions, hardcodes buried in formulas, and numbers that don't tie, and then fixes them.
Teams Plug-in
Model ML is now live as a plug-in in Microsoft Teams. It joins your chats like any other teammate, except this one already knows every project folder, source, and piece of context your team is working with. Tag @ModelML mid-conversation and it handles whatever you throw at it, whether it’s a quick research question, or an entire deck. And it learns as you go. If it spots something in the chat that needs doing, it can just run it for you.
GPT-5.6
Three new GPT-5.6 models have launched in Model ML: Sol, Terra and Luna. Sol is the most capable, Luna the fastest. Leave the picker on Auto and Model ML will pick the best model for you, or choose a specific model when you want to trade depth for speed.
Schedule and Event-Based Agents
Two big upgrades have landed on Schedules. You can now trigger a run after a given event rather than only at a set time, whether that's a new transcript being ready, a meeting about to begin, or a company on your watchlist reporting earnings. Now, you can also point a schedule at a specific agent to carry out the whole task automatically, with results delivered straight to your inbox or via webhook.
Homepage Redesign
Your work officially has a new home. Projects live in the sidebar, chat home works across all your projects, and everything you make is one click away. More refinements are landing weekly.
Model ML MCP
You can now use Model ML from your other AI tools. It plugs into ChatGPT, Copilot, Claude, Cursor, Gemini, and other MCP-compatible clients, so you can call the agent straight from the tools you use every day.
Multi-file Skills
Capture your most complex procedures with a single skill. It now holds the instructions, reference files, and variants for different cases, structured so the agent navigates straight to what each step needs.
Claude Opus 5
Claude Opus 5 is now live in Model ML. We tested it pre-release across the workflows our clients run every day, and its performance was consistently strong. The trade-off is cost. Opus 5 is one of the most expensive models to run, and GPT-5.6 Sol matches it on many tasks using up to 66% fewer tokens. This is precisely why we stay model agnostic. Since no single model leads on both price and performance, Model ML routes each task to the model best suited to it.
Custom MCP
Because Model ML adapts to how you work, you can now connect your own data sources and tools, then query them right in chat. Add any MCP server with a single OAuth sign-in, and your agent works across all of them alongside everything else it knows.
