Benchmark Registry

Coverage

Recorded results across recent models and the most covered benchmark families. Empty cells indicate gaps in this registry.

Updated

Recent models are selected in rounds across providers, up to 25 models total. Cells link to the newest reported observation in each family.

Coverage matrix

ModelSWE-benchHumanity's Last ExamTerminal-BenchGraduate-Level Google-Proof Q&ADeepSWEOSWorldAutomationBenchBrowseCompSWE-bench ProMCP AtlasMMMU-ProCharXivMMLU-ProCursorBenchIFBench
Gemini 4 ArgonGoogle (DeepMind)——Reported—ReportedReportedReported————————
GPT-6.1 SolOpenAI——Reported—ReportedReportedReported————————
Claude Sonnet 5.5AnthropicReportedReportedReported—Reported—Reported—Reported————Reported—
Grok 4.7SpaceXAI——Reported—Reported————————Reported—
Muse Spark 1.3Meta AI (originally Facebook AI Research)——Reported—ReportedReported—————————
GLM-5.3-FlashZ.ai—ReportedReported—Reported—Reported————————
MAI-Thinking-1Microsoft AIReported—ReportedReported————Reported———Reported—Reported
Qwen3.8-MaxAlibaba (Tongyi)—ReportedReportedReportedReportedReported————————Reported
Inkling-SmallThinking Machines LabReported—Reported————ReportedReportedReported—Reported——Reported
Kimi K3Moonshot AI—ReportedReportedReportedReportedReported—Reported—ReportedReportedReported———
NVIDIA Nemotron 3 Ultra 550B-A55BNVIDIAReportedReportedReportedReported———Reported—Reported—————
MiniMax M3MiniMaxReported—Reported————ReportedReported——————
Mistral Medium 3.5Mistral AIReported——————Reported——————Reported
Composer 2.5CursorReported—Reported——————————Reported—
DeepSeek-V4-FlashDeepSeek AIReportedReportedReportedReportedReported——ReportedReported———Reported——
Claude Opus 5.5AnthropicReportedReportedReported—Reported—Reported—Reported————Reported—
GPT-6 LunaOpenAI————ReportedReportedReported————————
Gemini 3.8 FlashGoogle (DeepMind)——Reported—Reported——————Reported———
GLM-5.3Z.ai——Reported—Reported—Reported——Reported—————
Grok 4.6SpaceXAI————Reported————————Reported—
Muse Glimmer 30BMeta AI (originally Facebook AI Research)Reported—ReportedReported—Reported——Reported—ReportedReported——Reported
InklingThinking Machines LabReported—ReportedReported———ReportedReportedReported—Reported——Reported
Qwen3.7-PlusAlibaba (Tongyi)Reported—ReportedReported—Reported——Reported———Reported—Reported
MAI-Code-1-FlashMicrosoft AIReported—ReportedReported————Reported——————
DeepSeek-V4-ProDeepSeek AIReportedReportedReportedReportedReported——ReportedReported———Reported——

By provider

ProviderModelsBenchmarks coveredRecords
Alibaba (Tongyi)82780
Anthropic1621199
Cursor4315
DeepSeek AI21054
Google (DeepMind)1637120
Meta AI (originally Facebook AI Research)52666
Microsoft AI31419
MiniMax3815
Mistral AI51023
Moonshot AI31745
NVIDIA3720
OpenAI1624214
SpaceXAI3612
Thinking Machines Lab21123
Z.ai61340

Stale results

Benchmarks whose newest reported result is more than 90 days before the latest data timestamp.