News
Updates, product notes and practical coverage for AI platforms and operators.
NVIDIA has released Nemotron 3.5 Super VL, an open and customisable multimodal mixture-of-experts model for agents working across text, images, documents and video. The 120-billion-parameter system activates 12 billion parameters per token. NVIDIA claims strong single-GPU throughput and highlights customer post-training gains, which adopters should reproduce on their own workloads.
DeepSeek has launched V4.1-Flash, a native multimodal model now live on its API. The company says the 552-billion-parameter mixture-of-experts model improves speed and cost, adds visual understanding, introduces lower pricing, and will temporarily route V4-Pro requests to the new model. Developers should test the planned routing change before it reaches production workloads.
ByteDance Seed has released SeedRealtime, a full-duplex model that continuously processes audio, video and text while deciding when to speak or act. The model is designed for fluid, interruption-aware interaction, with demonstrations spanning noisy group conversations, visual guidance and proactive assistance. Access and pricing details remain limited.
MiniMax has launched H3, an omni-modal generation model that accepts text, image, video and audio context and produces video with native stereo sound. The service supports clips up to 15 seconds at up to 2K resolution, with open model weights promised after release checks.
Black Forest Labs has opened early access to FLUX 3, a multimodal foundation model spanning image, video, audio and action prediction. Video generation is available first, while image, robotics and open-weight components remain on a staged release plan.
Baseten has added day-one access to Thinking Machines Lab’s Inkling model through its managed Model APIs and Dedicated Inference service. The launch gives developers a hosted route to a very large open-weight, multimodal model, while leaving performance, cost and production suitability to be tested in real workloads.