AI Product Updates: New Features and Releases

Zenith Studio positions itself as model-agnostic, letting users pick different image and video models in one place as needed.

The new image editor supports expand, re-angle and relight, while the video agent can create, edit, narrate and score a full video from a prompt or reference.

Factory is now Zenith Studio, combining image and video models, result editing, agent-built full videos and direct publishing to social platforms.

Fish Audio's S2.1 Pro voice model is now available on Runway for generating expressive voiceovers.

Plane Agents are now available: assign them a work item, tag them in comments, or run them on a schedule.

Z.AI's GLM-5.3-FlashX is live on B.AI's API and web chat, a 320B total/18B active sparse model with a 1M context window, native multimodality and up to 200 tokens/sec for interactive coding and agent workflows.

Daniel Jeffries' open-source project gives each person a long-term-memory Genie agent, lets people and agents share rooms, and ships co-editable apps for docs, video, design and terminal.

Honcho 3.2 lets chat calls pass include_evidence=True to return the conclusions the agent read, the messages it referenced and the tools it called, off by default and at no extra model token cost.

Cloudflare Python Workers are now generally available, supporting FastAPI, Django and Flask, with direct access to R2, D1 and queues and the ability to run MCP and RAG workflows.

Hugging Face has released Gradio Workflow, which lets users build AI pipelines by connecting nodes.

The local-first Hermes Agent has been updated to v0.11.0, adding Google Gemini integration, MCP SDK 2.x support, direct bot chat and stricter security, with 137 tools.

DopaMint has launched on Base, turning natural-language requests into onchain and real-world execution through chat, iMessage or its own app.