Runware

Gemini Omni Flash 1.1 is now on Runware 🎬 Google's Omni Flash 1.1 update adds 4K resolution, scene extension, and finer control over outputs. Extend clips up to 30 seconds by adding 3 to 10 second segments, define a start and end frame for interpolation, and pass reference videos alongside your prompt to keep character design consistent. Resolution options run from 360p through 720p, 1080p, up to full 4K. Launch link in the comments.

Recraft V4 Styles now on Runware 🎨 Build a visual style from reference images and generate new assets in that style. The V4 Styles family keeps texture, rendering, and composition consistent across generations, without training or LoRAs. Four models cover raster and vector output at two quality tiers; Recraft V4 Styles and Styles Pro handle raster for branded content, product campaigns, and editorial work, with the pro model stepping up detail retention for output that needs to be presentation-ready. Recraft V4 Styles Vector and Recraft V4 Styles Pro Vector bring the same workflow to SVG, suited to icon libraries, brand systems, and packaging. Launch links in the comments.

Wan 3.0 Prime is now available on Runware ⚡ Prime is Wan 3.0's faster inference tier, retaining the API shape, multimodal creation & editing workflows, and output quality. If you're already using Wan 3.0, switching to Prime is a simple model ID swap. The full capability set is available: text-to-video, first-frame and first-and-last-frame image-to-video, reference-driven generation with up to 10 reference images, 5 reference videos, and 5 reference audio inputs, native audio, and the model's 30-second native outputs. If you need Wan 3.0 quality in a faster package, switch to Prime now. Launch link in the comments.

Wan 3.0 is now on Runware on D0 🎬 Alibaba's new all-in-one multimodal video model. Text-to-video, first-frame and first-and-last-frame image-to-video, reference-driven generation, and audio-led creation. The reference capacity especially stands out: up to 10 reference images, 5 reference videos, and 5 reference audio files per request, generating 30-second clips, with native audio, and an optional thinking mode for complex multi-input prompts and storyboard references. Built for branded storytelling, product demos, explainers, and music-led sequences. Launch link in the comments.

Open-weight Ideogram 4.0q & Background Remover now on Runware 🎨 Ideogram released the weights for their 9.3B design-focused image model. The 4.0q variant is a quantized build, now accessible on Runware via the Playground and API. The model retains the design capabilities from 4.0: bounding-box layout control for placing text and elements at exact coordinates, structured JSON prompting for reproducible output, color-palette guidance, and multilingual text rendering. LoRA support is included as well, which makes it a solid base for building custom style workflows around Ideogram 4's core design strengths. Ideogram Background Remover is also available, producing clean transparent PNG cutouts with strong edge fidelity. Launch links in the comments.

FLUX Video Upscale is now on Runware. Black Forest Labs built their reputation on image quality and fidelity. FLUX Video Upscale brings that expertise to video enhancement. FLUX Video Upscale is a FLUX 3 powered super-resolution upscaler. Pass in a clip, pick your upscale factor, and get crisp results up to 4K. Two modes give you control over how the model handles detail; Precise mode sharpens without touching the source, important for footage with faces, products, or brand assets. Creative mode recovers and invents fine detail more aggressively, and works well on generated footage, textures, and scenery. Your audio track carries through automatically, and an optional text prompt lets you steer the details creative mode restores. Launch link in the comments.

Seedance 2.5 now generates in native 1080p on Runware 🎬 Full HD with native 10-bit color depth output is live. No post-generation upscaling required, just production-ready HD video straight from the API. If you're already running Seedance 2.5 on Runware, the upgrade is available now. Link in the comments.

We're the top performer in latency on Vercel Gateway along with Baseten but we're 40% cheaper 🫰 for Deepseek V4 Flash! We're also top 3 by throughput - higher than Fireworks AI or Deepseek. Can't wait to roll out more LLMs to your favorite gateway. It's been great to work with the Vercel team on this! 💪 🚀

We're live on Vercel's AI Gateway. On DeepSeek V4 Flash 0731, we're currently top performer in latency and throughput, achieving P50 ttft of 0.8s and 200+ tokens per second. We're also the cheapest provider on that model at $0.08/M input and $0.15/M output, around 40% below DeepSeek's own API. No training on your data, zero data retention available on enterprise. One model on the gateway today. More to come soon. And 35+ more LLMs on our own API. Links in the comments.

Wonder 3.5 is now on Runware ✨ Topaz Labs' latest generative upscaler doesn't just sharpen pixels, it rebuilds what's actually missing. Skin, hair, fabric, entire textures get regenerated as the image scales up, which makes a real difference on compressed or low-res sources that would otherwise fall apart under a normal upscale. Structured content holds up too. Tables, charts, and dense blocks of text stay legible instead of turning to mush. And you're not locked into one setting: strength goes from Low to High depending on how much you want the model inventing detail versus sticking close to the source. Launch link in the comments.