DeepSeek Vision Model Achieves 2 Wins and 2 Losses Against Opus 4.8

The DeepSeek vision model V4-Flash-Vision-Exp demonstrated superiority over Opus 4.8 in two categories across four multimodal agent evaluations. In the Agents' Last Exam, V4-Flash-Vision-Exp scored 27.3, surpassing Opus 4.8's score of 25.7, and also outperformed it in ZeroBench with a score of 35.0 compared to Opus 4.8's 34.0. However, in ApexBench, V4-Flash-Vision-Exp scored 36.5, falling short of Opus 4.8's 39.4, and in Chartography, it scored 64.3, which is lower than Opus 4.8's 65.0. Compared to the previous text model V4-Flash-0731, V4-Flash-Vision-Exp saw an increase in ApexBench scores from 26.2 to 36.5, and in the Agents' Last Exam from 25.2 to 27.3. DeepSeek explained that V4-Flash disregards multimodal content, and the score improvements are interpreted as a result of its ability to process image inputs. In six out of seven text evaluations, V4-Flash-Vision-Exp scored higher than V4-Flash-0731, with the DeepSWE score rising from 54.4 to 59.3. In Toolathlon, V4-Flash-Vision-Exp scored 75.9, just 0.3 points behind Opus 4.8's 76.2. This benchmark represents DeepSeek's own measurements, expanding options for developers in multimodal model selection. The current figures confirm that the DeepSeek vision model has achieved 2 wins and 2 losses against Opus 4.8 across four multimodal agent evaluations.

This content is provided for general informational purposes only and doesn't constitute financial, investment, legal, or tax advice. Any events, rewards, online promotions, or related information mentioned herein should not be considered a recommendation, solicitation, or invitation to purchase, sell, trade, or otherwise deal in any crypto assets. Crypto assets are highly volatile and may result in loss. The availability of WEEX services, products, and related events may vary by region. You are responsible for ensuring that your participation is in accordance with applicable local laws and regulations.