Luma Labs opens Uni-1.1 API

Luma Labs has launched its Uni-1.1 API, a multimodal model offering text-to-image and editing features starting at $0.0404 per 2048 px generation.

The REST interface provides access to Uni-1, Luma's multimodal model (decoder-only autoregressive transformer), presented as capable of interpreting creative intent, composition, and spatial logic before generating pixels. Two endpoints structure the API (reasoning and generation), accompanied by Python, JavaScript/TypeScript, and Go SDKs, and a CLI, with no waitlist.

Use cases cover text-to-image and natural language image editing, with up to nine references per generation to ensure consistency of characters, products, styles, or brand identities. Luma claims training conducted in collaboration with filmmakers, VFX artists, and international creators, which results in native multilingual rendering (Chinese, Japanese, Arabic), a sensitivity to visual registers (manga, fashion, architecture), and fine-tuned management of lighting and materials.

In terms of performance, the studio highlights latency and cost half that of comparable models, with a 2048 px image generated in approximately 31 seconds. Uni-1.1 ranks in the top 3 of the Image Arena (text-to-image and image edit) and leads the Human Preference Elo ranking. Two pricing plans coexist: a usage-based Build plan (starting at $0.0404 per 2048 px image for Uni-1.1, around $0.10 for the Max version) and a guaranteed throughput Scale plan designed for production workloads, with SLA and dedicated support. The model remains freely testable at app.lumalabs.ai before switching to the API.