Inkling 2 Small on Terminal-Bench 2.1Thinking Machines removed Inkling 2 Small’s 85.1% on Terminal-Bench 2.1 with the Pi harness from its model card.
Inkling 2 Small on SWE-bench VerifiedThinking Machines removed Inkling 2 Small’s 82.2% on SWE-bench Verified from its model card.
Inkling 2 Small on BrowseCompThinking Machines changed the label on Inkling 2 Small’s 90.2% on BrowseComp.
Inkling 2 Small on ChartographyThinking Machines’ Inkling 2 Small post and model card printed 80.4% for Inkling 2 Small on Chartography with Python in Frontier Log’s copies of Nov 7, 2026.
LongCat-2.5 on Humanity’s Last ExamMeituan changed the footnote on LongCat-2.5’s 63.1 on Humanity’s Last Exam with tools.
Inkling on Humanity’s Last ExamThinking Machines’ model card for Inkling printed 30.0% on Humanity’s Last Exam, text only, on the day of release.
Inkling on GPQA DiamondThinking Machines’ model card for Inkling printed 87.9% on GPQA Diamond on the day of release.
Claude models on OSWorld-VerifiedAnthropic re-ran older models on OSWorld-Verified after fixing a bug in its zoom tool and raising the limit on tokens per turn from 16K to 128K.