Deepseek's V4-Flash-Vision matches Opus 4.8 for 1% of the cost

deepseek released v4-flash-vision-exp, extending its v4-flash model with image processing capabilities to support multimodal agent workflows. the model achieves performance close to opus 4.8 on internal benchmarks while offering specific technical advantages like processing up to 600 images per request and auto-normalizing inputs. developers can implement it using standard apis or the updated harness framework, with pricing adhering to the base model's rates. this release signals continued aggressive price-performance competition in the multimodal agent space.