DeepSeek V4 Flash Vision Multimodal Model

DeepSeek released DeepSeek V4 Flash Vision EXP, a separate multimodal model. Its multimodal agent capability is claimed to approach Opus 4.8, while text-only performance matches DeepSeek V4 Flash. Pricing remains consistent with the V4 Flash model.
Key insights
DeepSeek's new vision model offers competitive multimodal performance at unchanged pricing
The V4 Flash Vision EXP model is positioned as nearing Opus 4.8 in agentic multimodal tasks, while maintaining the text capabilities and cost of the existing V4 Flash. This makes it an attractive option for developers needing vision capabilities without a price increase.