ByteDance Seed
Seed Diffusion Preview is a large-scale, code-focused language model that uses discrete-state diffusion to generate code non-sequentially, achieving dramatically faster inference without sacrificing quality by decoupling generation from the token-by-token bottleneck of autoregressive models. It combines a two-stage curriculum, mask-based corruption followed by edit-based augmentation, to robustly train a standard dense Transformer, striking a balance between speed and accuracy and avoiding shortcuts like carry-over unmasking to preserve principled density estimation. The model delivers an inference speed of 2,146 tokens/sec on H20 GPUs, outperforming contemporary diffusion baselines while matching or exceeding their accuracy on standard code benchmarks, including editing tasks, thereby establishing a new speed-quality Pareto frontier and demonstrating discrete diffusion’s practical viability for real-world code generation.
Learn more
Mirage by Captions
Mirage by Captions is the world's first AI model designed to generate UGC content. It generates original actors with natural expressions and body language, completely free from licensing restrictions. With Mirage, you’ll experience your fastest video creation workflow yet. Using just a prompt, generate a complete video from start to finish. Instantly create your actor, background, voice, and script. Mirage brings unique AI-generated actors to life, free from rights restrictions, unlocking limitless, expressive storytelling. Scaling video ad production has never been easier. Thanks to Mirage, marketing teams cut costly production cycles, reduce reliance on external creators, and focus more on strategy. No actors, studios, or shoots needed, just enter a prompt, and Mirage generates a full video, from script to screen. Skip the legal and logistical headaches of traditional video production.
Learn more
Ray3.14
Ray3.14 is Luma AI’s most advanced generative video model, designed to deliver high-quality, production-ready video with native 1080p output while significantly improving speed, cost, and stability. It generates video up to four times faster and at roughly one-third the cost of its predecessor, offering better adherence to prompts and improved motion consistency across frames. The model natively supports 1080p across core workflows such as text-to-video, image-to-video, and video-to-video, eliminating the need for post-upscaling and making outputs suitable for broadcast, streaming, and digital delivery. Ray3.14 enhances temporal motion fidelity and visual stability, especially for animation and complex scenes, addressing artifacts like flicker and drift and enabling creative teams to iterate more quickly under real production timelines. It extends the reasoning-based video generation foundation of the earlier Ray3 model.
Learn more
Seedance
Seedance 1.0 API is officially live, giving creators and developers direct access to the world’s most advanced generative video model. Ranked #1 globally on the Artificial Analysis benchmark, Seedance delivers unmatched performance in both text-to-video and image-to-video generation. It supports multi-shot storytelling, allowing characters, styles, and scenes to remain consistent across transitions. Users can expect smooth motion, precise prompt adherence, and diverse stylistic rendering across photorealistic, cinematic, and creative outputs. The API provides a generous free trial with 2 million tokens and affordable pay-as-you-go pricing from just $1.8 per million tokens. With scalability and high concurrency support, Seedance enables studios, marketers, and enterprises to generate 5–10 second cinematic-quality videos in seconds.
Learn more
