WHAT NEW IN X10.29 THU

AI collections from X · 22 collections · 60,636 posts

Back to No.008 Gemini 4 Pro / Flash: benchmarks, hands-on tests and leaksUpdated 8 hours ago
gemini-4IMAGE

Gemini 4 Argon early benchmarks: still behind on coding and science

Google unveiled Gemini 4 Argon for complex workflows, cybersecurity and coding, with internal teams already using it for software and quantum tasks, but early benchmarks show GPT-6 Astra still ahead in frontier coding and science, while Opus 5.5 holds the lead in terminal-based coding evaluations.

Sam Badawi
@Sam_BadawiArchived 28 days agoOriginally posted 2026.10.01
Gemini 4 Argon early benchmarks: still behind on coding and science
Sam Badawi @Sam_BadawiOriginally posted 2026.10.01
Open original on X

$GOOGL Google unveiled Gemini 4 Argon, targeting complex workflows, cybersecurity and coding, with internal teams already using it for software and quantum tasks. Early benchmarks are mixed, with GPT-6 Astra still ahead in frontier software coding and science. Claude Opus 5.5 also retains the lead in terminal-based coding evaluations.

x.com/sam_badawi/status/2105388933756694962Archived 2026.10.01 10:02
Open original on X