AiRecMark/Intelligence/Performance leaders
C-03 · Performance leadersarchive-recorded · snapshot 2026-09-17

Performance Leaders: Top 25

The 25 highest performance scores (score.dims.performance) across all 448 archives, each shown against its overall score so the deviation is visible. Aggregation is a plain sort of recorded values — no re-weighting; method and N are published below.

Key numbers

Archives ranked
448
data/tools · N=448
Performance median
80
score.dims.performance · N=448
Top score
90
score.dims.performance · N=448
Leader
Cartesia
score.dims.performance · N=448
Correlation with overall
r=0.86
pearson(dim, overall) · N=448

Performance — top 25 of 448

sort: dims desc, then overall desc, then slug; deviation = dim − overall

#ToolCategoryPerformanceOverallDeviation
1CartesiaAudio & Voice9085.8+4.2
2PiperAudio & Voice9081.1+8.9
3PlayHTAudio & Voice8989.2-0.2
4FLUXDesign8988.3+0.7
5CursorCoding8889.2-1.2
6ZedCoding8886.1+1.9
7DeepL WriteWriting8885.8+2.2
8SupermavenCoding8881.4+6.6
9ChatGPTResearch8690.9-4.9
10ClaudeResearch8688.9-2.9
11PerplexityResearch8688.4-2.4
12Semantic ScholarResearch8687.5-1.5
13OpenAlexResearch8687-1
14FathomWorkflow Automation8686.1-0.1
15QuillBotWriting8685.2+0.8
16DeepgramAudio & Voice8685.1+0.9
17AssemblyAIAudio & Voice8684.5+1.5
18TavilyResearch8684.5+1.5
19Kagi AssistantResearch8683.9+2.1
20ExaResearch8683.2+2.8
21GrammarlyWriting8683.1+2.9
22Adobe FireflyDesign8682.8+3.2
23PhindResearch8682.8+3.2
24Wispr FlowAudio & Voice8682.8+3.2
25Kokoro TTSAudio & Voice8682.5+3.5

Performance — top 10

Cartesia90Piper90PlayHT89FLUX89Cursor88Zed88DeepL Write88Supermaven88ChatGPT86Claude86

Recorded score.dims.performance of the ten leading archives (scale 0–100).

Method

  • Population: all 448 archives in data/tools with a recorded score.dims.performance.
  • Ordering: dimension score descending; ties broken by overall descending, then slug ascending — a deterministic total order.
  • Deviation column = score.dims.performance − score.overall per archive; positive values lead their own overall.
  • Pearson r between the dimension and overall is computed over all 448 pairs (standard formula, rounded to 2 decimals).
  • Scores are quoted as recorded in each archive; nothing is recomputed or re-weighted.
Data appendix. Source: data/tools/*.json · snapshot 2026-09-17 · method: sort(score.dims.performance desc, overall desc, slug) top 25 + deviation + pearson · aggregates published with method and N. T1 benchmarks recorded: 3 / 448.