Black Forest Labs’ FLUX 3 trains image, video, audio, and robot actions inside one unified architecture — but for now, you’ll need an invite to touch it.

Google’s Gemma 4 family spans 2.3B to 31B parameters, with an encoder-free 12B that reads raw audio — all Apache 2.0 licensed and phone-friendly.

Fish Audio raised $52M on $21M ARR and 8M users — but its open voice models raise cloning-misuse questions that no amount of funding can fix.