Frontier Log

Benchmark · Active

Chartography

Created by Surge AI, 16 Jul 2026 · Higher is better · Website

020406080100Oct 2025Jan 2026Apr 2026Jul 2026Oct 202690.2%Sonnet 5.585.2%Mythos 58.8%Haiku 4.5020406080100Oct 2025Jan 2026Apr 2026Jul 2026Oct 202690.2%Sonnet 5.585.2%Mythos 58.8%Haiku 4.5

Results

ModelScoreSourceSource check
Kimi K3.1 Pro24 Oct 2026 · Moonshot AI90.2%Kimi K3.1 Pro Tech Blog: Advancing Open IntelligencePython tool; max reasoning effort; average of 5 runs
DeepSeek-V4.1-Flash-101717 Oct 2026 · DeepSeek83.0%DeepSeek-V4.1-Flash-1017: Wish-Driven Programming Made RealSource image
CorroboratedThe same number is printed in 1 other lab document: 1
Claude Haiku 5.57 Oct 2026 · Anthropic86.2%Claude Haiku 5.5 System Cardadaptive thinking, max effort; container and image-cropping tool; average of 5 runs; Gemini 3.5 Flash grader
Claude Sonnet 5.528 Sep 2026 · Anthropic90.2%Claude Sonnet 5.5 System Cardadaptive thinking, max effort; container and image-cropping tool; average of 5 runs; Gemini 3.5 Flash grader
CorroboratedThe same number is printed in 1 other lab document: 1
Claude Opus 5.522 Sep 2026 · Anthropic89.0%Introducing Claude Opus 5.5adaptive thinking, max effort
CorroboratedThe same number is printed in 3 other lab documents: 1 2 3
DeepSeek-V4.1-Flash10 Sep 2026 · DeepSeek78.9%DeepSeek-V4.1-Flash: Smarter, Faster, More EfficientSource image
CorroboratedThe same number is printed in 1 other lab document: 1
Claude Fable 5.11 Sep 2026 · Anthropic86.2%Claude Fable 5.1 and Claude Mythos 5.1 System Cardadaptive thinking, max effort; container and image-cropping tool; average of 5 runs
Claude Fable 5.11 Sep 2026 · Anthropic88.4%Introducing Claude Opus 5.5reported later; another report, not charted
CorroboratedThe same number is printed in 3 other lab documents: 1 2 3
GLM-5.3-Flash26 Aug 2026 · Z.ai78.0%GLM-5.3-Flash: Frontier Intelligence, Flash Cost256K contextOur capture
Claude Opus 524 Jul 2026 · Anthropic83.0%Claude Opus 5 System Cardadaptive thinking, max effort; container and image-cropping tool; average of 5 runs
CorroboratedThe same number is printed in 1 other lab document: 1
Claude Opus 524 Jul 2026 · Anthropic83.4%Introducing Claude Opus 5.5reported later; another report, not charted
CorroboratedThe same number is printed in 1 other lab document: 1
Claude Sonnet 530 Jun 2026 · Anthropic71.8%Claude Opus 5 System Cardadaptive thinking, max effort; container and image-cropping tool; average of 5 runs; reported later
Claude Sonnet 530 Jun 2026 · Anthropic73.2%Claude Opus 5.5 System Cardadaptive thinking, max effort; container and image-cropping tool; average of 5 runs; Gemini 3.5 Flash grader; reported later; another report, not charted
Claude Fable 59 Jun 2026 · Anthropic84.2%Claude Fable 5.1 and Claude Mythos 5.1 System Cardadaptive thinking, max effort; container and image-cropping tool; average of 5 runs; reported later
Claude Mythos 59 Jun 2026 · Anthropic85.2%Claude Opus 5 System Cardadaptive thinking, max effort; container and image-cropping tool; average of 5 runs; reported later
Claude Opus 4.828 May 2026 · Anthropic75.0%Claude Opus 5 System Cardadaptive thinking, max effort; container and image-cropping tool; average of 5 runs; reported later
CorroboratedThe same number is printed in 1 other lab document: 1
Claude Haiku 4.515 Oct 2025 · Anthropic8.8%Claude Haiku 5.5 System Cardextended thinking, budgets up to 32K; container and image-cropping tool; average of 5 runs; Gemini 3.5 Flash grader; reported later

All benchmarks