Questions on DeepSeek V4 Pro Image Input
Porn197617_ · reddit · 2026-07-13
While building an AI writing agent workflow, the author switched the primary model from GPT-5.5 to DeepSeek V4 Pro and found that its API seemingly cannot directly read images.
Their usual workflow involves feeding design drafts, UI screenshots, or error screenshots to the agent so it understands the goal before proceeding. Without image capabilities, many layout and visual details must be described in text, significantly dropping efficiency. The author wants to confirm:
- Whether the DeepSeek V4 Pro API currently supports image input
- If not, should they use a vision model for image understanding first, and then pass the results to DeepSeek for subsequent tasks
- Whether there is an agent framework that can automatically select the appropriate model for different tasks without manual switching
More from coding & agent
- Soft Clamp cuts tool-call overuse in multi-teacher distillation, from 13.7% to 9.0% — antgroup · 2026-07-21
- Agent harness memory loss and compaction are still a major usability problem — adityaag · 2026-07-21
- SpecJudge runs locally on Ollama to pick the right-sized AI model for your project — jokiruiz · 2026-07-21
- A developer maps out six design rules for CLIs that humans and AI agents can both use — yujiezha · 2026-07-21
- A coding-agent skill that forces ADHD-friendly, answer-first output — ayghri · 2026-07-21
- A set of agent skills for CAD, robotics, and hardware design — earthtojake · 2026-07-21