Unverified50% confidenceBenchmarkExact time
The human baseline on the OSWorld benchmark is about 72.4%, and Mage's 66% is close to that level, surpassing a model nearly 6 times its size (63.9%).
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
10/31/2026
First Seen
Valid until: 1/29/2027
Sources
OpenCode Responds to Model Weakening Controversy; Microsoft Open-Sources 4B Desktop Agent
bilibili阿梨Aria早鸟报10/31/2026
Related Entities
Related Claims
UnverifiedWhen OSWorld was released in 2024, the strongest baseline model score was only approximately 12%, while the human completion rate was approximately 72%.74% similarUnverified人类在OS World测试中的平均得分约为72-74%67% similarUnverifiedThe new model achieved 1,620 points on GDPVal (compared to 735 for the previous generation), improved on OSWorld from 15.7% to 72.4%, on Terminal Bench from 0.0% to 39.2%, and on Humanity's Last Exam from 10.2% to 45.9% without tools and from 18.7% to 57.4% with tools.61% similarUnverifiedEven after fine-tuning to reach 65.7% accuracy, there is still a considerable gap from the human level of 87.2%.60% similarUnverifiedAmong the 21 models evaluated on SpaceCast-Bench, the strongest model reaches only 58.0% accuracy, while human performance is 87.2%.57% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1120152API
curl https://kongchang.com/api/v1/knowledge/claims/1120152MCP
get_claim(id=1120152)