Meta ships SAM 3.1 on Model API: noun-phrase prompts for detection, segmentation and tracking at $2.50/1k images
armand_ruiz · x · 2026-09-19
- Meta's Segment Anything Model 3.1 is now live on the Meta Model API, an official implementation with inference tuned for the model's architecture.
- A noun-phrase prompt finds objects in images and video; one API call returns detections, pixel-precise masks, and identity-preserving video tracks, sharing the same API surface, billing and auth as other Muse models.
- Pricing is $2.50 per 1,000 images and $0.20 per 1,000 frames; open weights remain available on Hugging Face.
Related event: Meta Launches SAM 3.1 API for One-Shot Detection, Segmentation and Tracking(2 posts)→
More from Multimodal
- Creator Shares 30-Second Epic Solo Battle Video Made with Seedance 2.5, Full Prompt Included — LudovicCreator · 2026-09-19
- Midjourney v8.2 prompt trick: swap one word to reshape the whole image — tisch_eins · 2026-09-19
- Local AI music: YuE2 generates a full cover song on your own machine via Pinokio — cocktailpeanut · 2026-09-19
- Runway CEO: one person made this AI film of a true story, frame costs collapsed — jnack · 2026-09-19
- Meta ships SAM 3.1 on Model API: detection, segmentation, tracking in one call — AIatMeta · 2026-09-19
- Runway's Ruby model adds alpha channel support with one-step HDR conversion — runwayml · 2026-09-19