gpt-6IMAGE

ACE-Brain-0.6 9B embodied model beats GPT-6 Astra on 6 of 8 grounding tests

Daxiao Robotics released a Qwen3.5-9B-based embodied model that outperforms GPT-6 Astra on six of eight grounding benchmarks, averaging 74.53 vs 73.68 across seven public tests.

SuSu_酥酥👅

@NFT_ChenArchived 10 days agoOriginally posted 2026.10.19

ACE-Brain-0.6 9B embodied model beats GPT-6 Astra on 6 of 8 grounding tests (image 1 of 2)

ACE-Brain-0.6 9B embodied model beats GPT-6 Astra on 6 of 8 grounding tests (image 2 of 2)

SuSu_酥酥👅 @NFT_ChenOriginally posted 2026.10.19

Open original on X

💥炸了！国产9B具身模型掀桌子，六项测试反超GPT-6 Astra！中国大晓机器人今日发布视觉空间锚定增强型具身基模型 ACE-Brain-0.6，基于Qwen3.5-9B监督微调，主打把语言指令精准落到物体、位置和可交互区域。对比GPT-6 Astra： 🔹八项Grounding评测中六项超越Astra 🔹七项公开基准平均分 74.53（Astra 73.68，领先0.85分） 🔹Where2Place：76.55 vs 71.10 🔹RoboSpatial：74.82 vs 69.55 🔹RoboAfford：88.51 vs 84.94 🔹自建真实机器人场景RobotGrounding：94.13 vs 91.10 数据驱动跃迁：约百万条空间锚定样本，含94万张图像+270万组问答，数据规模每翻倍综合分平均提升约2.4分。 9B小模型在专项上掀翻旗舰，参数崇拜可以歇一歇了。 #具身智能 #机器人 #大晓机器人 #ACEBrain #GPT6Astra #具身模型

x.com/nft_chen/status/2112137524919631909Archived 2026.10.20 10:11