Blog
New AI models, explained and ready to try
Practical notes on the open-source models we host: what they are good at, their license, and how to run them in your browser.
-
MOSS-Image online: the open image generator that gets text right, no GPU needed
OpenMOSS's MOSS-Image is a 5B open image model that writes clean text inside images and follows long prompts. What it is, its Apache-2.0 license, and how to try it in your browser.
-
Phonon TTS online: try Ollama's new voice model on any computer
Gradium's Phonon 31M is the text-to-speech model Ollama just added, but only on Apple Silicon Macs. Try it in your browser on Windows or Linux, and see which Phonon models allow commercial use.
-
Unlimited OCR 2 online: turn scans and photos into Markdown without installing anything
Baidu's Unlimited OCR 2 reads text, tables, formulas and layout from a page photo and returns clean Markdown. What it does, its MIT license, and how to try it in your browser.
-
6 open-source text-to-speech models you can try online: Qwen3-TTS, Chatterbox and more
Qwen3-TTS, Chatterbox, Ming 3.0, Audio8, Phonon and Paradee compared: languages, voice cloning, licenses and speed. All six run in your browser, free to try, no GPU needed.
-
LFM3 1.7B vs Spark-X2.6 1.7B: try tiny local LLMs online before you download them
Two new 1.7B chat models small enough for a laptop or phone. Compare them side by side in your browser before pulling them into Ollama or LM Studio.
-
New open-source AI models this week (November 2026) and where to try them
Nanbeige4.3-33B, Apex 35B, Groot 19B, Unlimited OCR 2, Paradee, Audio8 and more from this week's AI news: what each one does and a link to run it online.