Unverified50% confidenceBenchmarkExact time
Artificial Analysis measured that Gemini 4 Argon generates approximately 62,000 output tokens per task, while GPT-6 Astra uses only approximately 27,000 to complete the same work.
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
10/11/2026
First Seen
Valid until: 1/9/2027
Sources
Related Entities
Related Claims
UnverifiedOn purely agentic coding and computer operation tasks, GPT-6.1 SoC still slightly outperforms Gemini 4 Argon in certain specific scenarios.73% similarUnverifiedIn the Deep Study software engineering test, Argon scored 77.9%, higher than GPT-6 Astra and Claude Opus 5.5 (both approximately 74%).71% similarUnverifiedOpenAI states that the work was driven by an internal model claimed to far surpass its newly released GPT-6 Astra, and that the team solving the Navier-Stokes problem involved approximately 10,000 concurrent agents.71% similarUnverifiedAccording to Artificial Analysis data, Haiku 5.5 consumes approximately 162 output tokens per intelligence index task in max effort mode, roughly 3 times that of GPT-6 Luna in the equivalent mode.69% similarUnverifiedThe testing team invested more than 20 billion (20B+) tokens in a comprehensive exploration of GPT-6 Astra.67% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/997662API
curl https://kongchang.com/api/v1/knowledge/claims/997662MCP
get_claim(id=997662)