This release supercharges your creative workflows with Anthropic's most advanced AI model, a brand-new multi-model generation mode for comparing outputs side by side, and a suite of quality-of-life upgrades that give power users fine-grained control over how their workflows run. Plus, video nodes get a major usability boost with inline playback and frame extraction.

📦 New

• Opus 4.6 — Anthropic's most intelligent model, now available on FLORA. Brings deeper reasoning, stronger instruction-following, and more nuanced creative output to your text and multimodal workflows — ideal for complex chains that demand precision.

🧠 Improved

• Option + Drag to Duplicate — instantly clone any node by holding Option (Alt) and dragging, cutting out extra clicks when you're building or iterating on complex workflows

• Lock Node — pin a node's outputs and assets so they won't re-run when you execute the rest of the workflow, perfect for preserving hero assets while experimenting downstream

• Copy Prompt — right-click any node to copy its prompt straight to your clipboard for quick reuse or reference

• Video Playback Controls — all video nodes now include a scrubber and full playback controls so you can review output without leaving the canvas

• Extract Current Frame — grab any frame from a video node as a still image, making it easy to pull reference frames or create thumbnails directly in your workflow

• Refund Info on Failed Tasks — the activity queue now shows credit refund details when a task fails, so you always know where your credits stand

🔄 Changed

• Styles Sunset (Temporary) — the styles trainer, sidebar access, and style references on models have been removed for now. In practice, state-of-the-art models like Nano Banana Pro already deliver more adherent stylistic results through prompting alone. We're rebuilding styles from the ground up — expect more powerful capabilities like References and Elements for character and style consistency in an upcoming release. In the meantime, you can still upload LoRAs for supported models (Flux Dev, with Flux 2 and Qwen coming soon).