Fireworks AI released DeepSeek-V4-Flash-Vision-Exp, an experimental multimodal model that adds image understanding to the text‑focused DeepSeek-V4-Flash. It keeps the same speculative decoding architecture and can be run on dedicated GPUs through Fireworks' on‑demand service, which has no rate limits. The company says serverless access will end on September 25, 2026, but on‑demand deployment will remain.
Why it matters
Developers can now build agents that process both text and images using Fireworks' reliable GPU service.
