vela-4
Vela 4
Flagship reasoning and code
Our flagship model for agents, software engineering, and long-context work. Configurable reasoning, tool use, and a 500k context window.
- Context
- 500k
- Input / 1M
- $2.00
- Output / 1M
- $6.00
Models
Eight public models. One endpoint surface. Pick flagship depth, latency, cost, image, or video.
vela-4
Flagship reasoning and code
Our flagship model for agents, software engineering, and long-context work. Configurable reasoning, tool use, and a 500k context window.
vela-fast
Low-latency production
The same family, tuned for interactive products. Sub-200ms first token on typical prompts with a lower unit cost.
vela-mini
Efficient at scale
A compact model for classification, extraction, and high-volume workloads where cost and speed matter more than depth.
vela-imagine
Text to image
Photoreal and editorial image generation from a natural-language prompt. Returns hosted URLs ready for your product.
vela-embed
Embeddings
Dense vectors for search, clustering, and retrieval. 256 dimensions.
vela-video
Cinematic video clips
Cinematic text-to-video on Helios. 720p, 1080p, or 4K. 6–60 s. Hosted MP4.
vela-video-fast
Shorter queue, same family
Same Vela Video family, scheduled on Edge VX60 first for faster turnaround. 6–60 s clips.
vela-audio
Speech in and out
Text-to-speech and transcription on the same key as chat.