GPT-6.1 Terra
Higgsfield's Supercomputer adds OpenAI's fast, low-cost model for writing, document and image reading, and multi-step app tasks.
Every Higgsfield update in one feed — release notes, features and changes, straight from the source.
223 updates
Higgsfield's Supercomputer adds OpenAI's fast, low-cost model for writing, document and image reading, and multi-step app tasks.
Higgsfield Playable builds a playable mobile game from a prompt, a gameplay idea or your own assets in one click; playable ads are for Enterprise.
Playable is open to everyone on web, mobile and Claude, while the Playable Ads export for ad networks is limited to Enterprise plans.
NVIDIA and TCV have joined Higgsfield’s Series B extension. Higgsfield says it now has over $1 billion in annualized revenue and 40 million users.
Moonshot AI's open-weight model is one of the lower-cost options in Supercomputer's selector, with a 1M-token context and text and image input.
Ideogram's SVG model turns prompts or a single image into editable vector logos and icons, with Low, Medium or High quality settings.
Any Higgsfield account, free or paid, can enter the 120-second film contest, with top prizes of 30,000 credits and winners announced each Wednesday.
Higgsfield’s new Content Factory makes ready-to-post reels, from “What if” explainers to animated hooks for ads. It’s available now.
Higgsfield now runs Kling 4.0, with Omni Reference for characters, objects and audio, start, middle and end frames, and native 4K coming soon.
Kling's latest video model on Higgsfield takes up to 15 references and 10 keyframes, edits finished clips, and makes 3 to 30 second videos.
The lighter Genjutsu drops your photo into any trending video, with generations costing 3.5x less at 720p and 5x less at 1080p.
Chinese is available to all Higgsfield users, switchable from the profile menu, Supercomputer or the footer language selector.
Higgsfield's ChatGPT plugin builds a 3D scene from a photo, video or text, with 24 to 300 mm lenses and video output up to 30 seconds.
The free-to-watch beginner course runs 1 h 42 min, and its required homework needs an active subscription and credits.
Ten $5,000 cash prizes go to Higgsfield entries about Iman Gadzhi's investment, with Instagram posts and Discord submissions open to entrants aged 18 and over.
The updated upscaler gives sharper detail in less time and for fewer credits, and it is currently available only on Higgsfield.
Anyone can browse creators' shots and open projects without an account, and Open Projects with a 1-minute video can earn up to 200,000 credits.
Pro plans and above get a free week of Wan 3.0 with 30-second 1080p clips, and API users get 60% off until October 31, 2026.
The Claude skill supports up to five reference images per edit and lets creators publish community presets using their own videos.
Paid accounts can cast up to five votes, and the film with the most votes wins $100,000.
Entries close October 14, with 200,000 credits in prizes and community voting open to accounts with an active paid plan.
The rollout spans Flow, ElevenCreative, fal, Magnific, OpenArt, Pika and Krea, with distinct FLORA and Weave announcements.
Reuse one reference video to change products, cast, locations, scripts, hooks and CTAs, with output up to 1080p.
Submissions remain closed, with the shortlist due October 8 and winners to be announced in the last week of October.
Credits for failed generations were refunded automatically, so affected users can retry at no extra cost.
Every Higgsfield user gets five free generations to try building a character and bringing it into an existing video.
FLUX 3 Image from Black Forest Labs adds text-to-image and image-to-image modes on Higgsfield, with native 2K and 4K rendering.
Higgsfield users can generate sound and motion together, use up to seven reference images, and extend clips from their last frame.
Ads Studio builds static ad batches from a website link, with brand settings saved once and reused, and up to 20 ads per product.
Ideogram 4.5 handles repeated edits, text translation and style transfer from up to 4 references, at 1K or 2K, from the Image model selector.