gpt-6IMAGE
BVB leaderboard: Astra xhigh tops at 73.8
In the updated BVB leaderboard, GPT-6 Astra xhigh leads with an overall score of 73.8 and is the only model to break 60 on Dual VQA, followed by Sol, Opus 5.5, and Luna.
Yolo Y. Tang
@yoloytangArchived 2026.09.30Originally posted 2026.09.29
Yolo Y. Tang @yoloytangOriginally posted 2026.09.29
📈 BVB leaderboard update! New results for Claude Opus 5.5 at high and xhigh, plus GPT-6 Astra, Sol, and Luna at xhigh. 🥇 GPT-6 Astra xhigh leads with an overall score of 73.8 and is the only model to break 60 on Dual VQA. 🥈 GPT-6 Sol xhigh takes second at 71.4 for just $0.40 per scene. 🧠 Claude Opus 5.5 xhigh reaches 69.8, gaining 4.1 points over high. 💸 GPT-6 Luna xhigh scores 68.1 for about $0.01 per scene, the lowest cost on the board. 🔗 Full leaderboard: https://yoloytang.me/BVB/#leaderboard
This post quotes another post: x.com/yoloytang/status/2099680399232516435
x.com/yoloytang/status/2104729605290778983Archived 2026.09.30 10:24