Frontier Log

Benchmark · InactiveSuperseded

SEC-bench Pro 260505

Created by University of Illinois Urbana-Champaign, 5 May 2026 · Higher is better · Website

020406080100Oct 202688.2%Gemini 4 Carbon62.8%DeepSeek-V4.1-Flash83.4%Gemini 4 Argon56.4%DeepSeek-V4-Pro-0813020406080100Oct 202688.2%Gemini 4 Carbon62.8%DeepSeek-V4.1-Flash83.4%Gemini 4 Argon56.4%DeepSeek-V4-Pro-0813

Results

ModelScoreSourceSource check
Gemini 4 Carbon28 Oct 2026 · Google88.2%Gemini 4 Carbon: our next frontier model183-task JavaScript engine split (May 2026); pass@1, mean of 3 attempts; Antigravity harness; LLM judge for crashes in both buildsSource image
CorroboratedThe same number is printed in 1 other lab document: 1
Gemini 4 Argon30 Sep 2026 · Google83.4%Gemini 4 Carbon: our next frontier model183-task JavaScript engine split (May 2026); pass@1, mean of 3 attempts; Antigravity harness; LLM judge for crashes in both builds; reported laterSource image
CorroboratedThe same number is printed in 1 other lab document: 1
DeepSeek-V4.1-Flash10 Sep 2026 · DeepSeek62.8%DeepSeek-V4.1-Flash: Smarter, Faster, More EfficientClaude Code harness; max reasoning effortSource image
CorroboratedThe same number is printed in 3 other lab documents: 1 2 3
DeepSeek-V4-Pro-081313 Aug 2026 · DeepSeek56.4%DeepSeek-V4.1-Flash: Smarter, Faster, More Efficientmax reasoning effort; reported laterSource image
CorroboratedThe same number is printed in 3 other lab documents: 1 2 3

All benchmarks