This week: Claude Opus 4.5 and Gemini 3 arrive, plus playgrounds get the upgrade they deserved.
Claude Opus 4.5 and Gemini 3
Claude Opus 4.5 is now available for orgs that have it enabled - the newest model from Anthropic. Early testers say it handles ambiguity and reasons about tradeoffs without hand-holding.
Google's Gemini 3 Pro Preview is also live. It significantly outperforms 2.5 Pro on every major benchmark with state-of-the-art reasoning and multimodal capabilities and a 1M context window.
Improved Prompts
We added a setting to hide the first user message when someone runs your prompt. If you've built a prompt with a long message it no longer clutters their chat. You can optionally show a custom message instead - for example just their input.
Playground Overhaul
Playgrounds - the side-by-side prompt testing feature - now stream results in real-time with stop buttons on each column and a "stop all" in the header. We also added MCP support per column (Anthropic and OpenAI models only), and file attachments that save with the playground so you're not re-uploading the same PDF every time you iterate.
Document Splitting in Workflows
We added a new workflow node that takes a PDF and outputs an array of files - one per page. Feed a 50-page document in, get 50 single-page files out, ready for iteration. This is the basis of all our ingestion strategies and is required when a document is too large to send to an LLM in one shot. Split it, loop through it, recombine results.
And More
- Workflow logs now paginate for large executions (100+ steps load progressively)
- You can type while responses are streaming - the input field no longer greys out
- Right-click menus cleaned up across the app
- Links in LLM responses now open in new tabs
Coming Soon
Prompt Assistant - a tool to help refine and optimize your prompts automatically. Also in the pipeline: parallel execution nodes, enhanced scraping nodes, Microsoft Outlook integration, and file uploads to Memories.