New AI APIs for assets, search, and video
Need something useful from the AI firehose? Start here: better image assets, broader agent research, a new coding model, and video cleanup you can call by API.
In today’s edition
- GPT-Image-2 can now generate transparent-background assets through the API preview.
- ChatGPT Work and Codex can reach Exa’s index of more than 100B websites, papers, and docs.
- OpenRouter released a free multimodal coding model with a 1M-token context window.
- FLUX Video Upscale is available through OpenRouter’s Video API in 2K and 4K.
- Claude Academy launched free courses and tutorials for people learning to use Claude.
GPT-Image-2 adds transparent backgrounds
GPT-Image-2 can now produce images with transparent backgrounds in API preview. That removes a tedious cleanup step when you need an asset to live in more than one place.
- Generate reusable visual assets that sit cleanly on any background.
- Use it for product imagery, graphic design, website mockups, and marketing campaigns.
- The feature is available in preview through the API.
What it means Test transparent asset generation in your creative pipeline so product images, ads, and site components can move from prompt to placement with less manual editing.
Source: OpenAIDevs
Exa brings web-scale research to Codex
ChatGPT Work and Codex can now use the Exa plugin to reach more than 100B websites, papers, docs, and other sources. Your research-heavy workflows just got a much bigger shelf to browse.
- The plugin gives ChatGPT Work and Codex access to Exa’s index.
- The stated coverage includes websites, papers, docs, and more.
- It targets work that needs outside information rather than only the files already in your project.
What it means Try the Exa plugin for research, competitive scans, and source gathering when your ChatGPT Work or Codex task needs information beyond its immediate context.
Source: OpenAIDevs
OpenRouter launches the free Ox Alpha model
OpenRouter introduced Ox Alpha, a stealth frontier model aimed at efficient coding, sustained agent work, and production use. It accepts text, image, and video input, with room for very long tasks.
- Ox Alpha has a 1M-token context window.
- It supports text, image, and video inputs.
- OpenRouter says the model is free to try.
- The provider does not train on prompts or completions, according to OpenRouter.
What it means Bench Ox Alpha against your current coding or agent model on long, multimodal jobs, especially if free access and prompt-data handling affect your evaluation plan.
Source: OpenRouter
FLUX Video Upscale lands on OpenRouter
FLUX Video Upscale from Black Forest Labs is now on OpenRouter, bringing 2K and 4K enhancement to the Video API. That gives video builders a direct route to sharper output without changing tools.
- Upscale videos to 2K or 4K through OpenRouter.
- The service targets sharper faces, cleaner textures, and finer background detail.
- It is designed to retain natural motion and realism.
- Choose `precise` for faster results and stronger identity retention, or `creative` for more repaired and generated detail.
What it means Add an upscale step after video generation when client delivery, social publishing, or product demos need higher-resolution output without rebuilding your video workflow.
Source: OpenRouter
Claude Academy opens free AI training
Claude Academy is live with free courses and tutorials open to anyone. Whether your team is new to AI or already using Claude daily, the program is built with paths for both.
- Courses and tutorials are free.
- The materials are open to anyone.
- Claude Academy includes paths for people starting with AI and people already using Claude every day.
What it means Use the free curriculum to give your team a shared starting point, then turn the most relevant lessons into repeatable internal workflows and onboarding material.
Source: claudeai