Snapshot 2026-09-21 · rebuilt daily

RL environments and AI benchmarks, cataloged daily.

105 RL environments. 3,144 AI benchmarks. 24,924 model scores. One page per entry: builder, domains, source, related work.

Browse environments Browse benchmarks Model leaderboard
105
RL environments
3,144
AI benchmarks
24,924
model scores
22
domains

Domain map

Environments and benchmarks by domain

Click a hub to list its entries. Click a circle (RL environment) or diamond (AI benchmark) to open its page.

Domain

Select a node to trace its connections. Zoom in for more space; drag to pan.

Biggest domains

Ranked by total entries. 22 domains in all.

01 Coding 67 36 env · 31 bench 02 Enterprise Workflows 33 25 env · 8 bench 03 Long-Horizon 26 22 env · 4 bench 04 Computer Use 25 20 env · 5 bench 05 Agentic 16 0 env · 16 bench 06 Knowledge 14 0 env · 14 bench
Show 16 smaller domains ↓ Hide smaller domains ↑

Today's signals · 2026-09-21

What changed in the catalog

September 21, 2026 was a near-static day. The sole change across all seven tracked datasets was a single leaderboard row added to Design Arena — React Native Apps, bringing its total to 46 rows. All other counts held flat: 3,144 benchmarks, 24,924 models, 1,702 companies, 105 environments, 10 tools, 66 occupations, and 1 research article.

Design Arena — React Native Apps Reaches 46 Rows

The multimodal mobile-app generation leaderboard gained one submission today, ticking from 45 to 46 rows. This marks continued incremental community engagement with Design Arena's mobile category, consistent with the steady row growth seen across Design Arena leaderboards throughout September.

The catalog

Environments, benchmarks, domains, and models

Each entry has its own page, and you can sort or filter any of the tables. The snapshot rebuilds daily.

80

RL environments

Training and eval environments for agents, with builder and domains.

Browse → 80

AI benchmarks

Public eval suites and leaderboards referenced across the ecosystem.

Browse → 22

Domains

Coding, computer use, enterprise workflows, and more, each with its entries.

Browse → 50

Model leaderboard

Ranked by BenchLM overall score, with creator and rank.

Browse →

Research

Latest analysis

Original data on the RL environment economy. All research →

  1. Laya vs Jev: Did Decision AI Arrive a Year Earlier?
  2. Is TypeSafe's Jev an LLM? What changes for RL environments
  3. Anthropic's AI agents beat 28 human experts at alignment research, and humans became the baseline