Directory label

Agentic benchmarks and RL environments

A label other catalogs put on entries, not one of the domains on our map. On the map, each entry here sits in the domain its own site or paper leads with.

Models acting as agents: taking actions in a terminal, browser or tool environment over several turns to reach a goal, rather than answering a single prompt.

AI benchmarks
477
Shares entries with
12 categories

Snapshot 2026-11-05. How categories are built · Where the data comes from

In short

What are the main agentic benchmarks?
Terminal-Bench 2.1, Tau2-Bench Telecom, Terminal-Bench 4.0, OSWorld, Terminal-Bench-Science 0.1 and AutomationBench, among 410 confirmed below.

Only benchmarks use the Agentic tag: the RL environment directories file their vendors under other categories. Browse all RL environments.

AI benchmarks 477

  1. Terminal-Bench 2.1
  2. Tau2-Bench TelecomSierra Research
  3. Terminal-Bench 4.0
  4. OSWorldUniversity of Hong Kong, Salesforce, Carnegie Mellon University, Waterloo
  5. Terminal-Bench-Science 0.1Stanford
  6. AutomationBenchZapier
  7. Finance Agent v2Vals AI
  8. AgentBench FCTsinghua University
  9. JobBench
  10. GDPvalOpenAI
  11. E-Bench
  12. DRACOPerplexity
  13. AA BriefcaseArtificial Analysis
  14. BankerToolBench
  15. BFCL v3UC Berkeley
  16. Finance Agent v1.1Vals AI
  17. ALE-BenchSakana AI
  18. GDPval-AAArtificial Analysis
  19. Gert Labs
  20. AndroidWorld
  21. MCP Mark VerifiedMoonshot AI
  22. PinchBenchKilo
  23. QwenClawBenchAlibaba
  24. Harvey's Legal Agent Benchmark
  25. Vending-Bench 2Andon Labs
  26. Tau3-BankingSierra Research
  27. Berkeley Function-Calling Leaderboard
  28. ClawEval-MM
  29. Gert Labs Rankings
  30. RuneBench
  31. AgentDrive
  32. Gaia2
  33. MCPMark-VerifiedMoonshot AI
  34. ProfBench
  35. SearchAgent Leaderboard
  36. Vals.ai Financial Agent 1.1
  37. Terminal-Bench-Science
  38. Long-Horizon Terminal-BenchTencent HY LLM Frontier
  39. GDP.pdfSurge AI
  40. ToolathlonHKUST NLP
  41. AutomationBench-AAZapier
  42. τ-benchSierra
  43. PostTrainBench
  44. DABstepAdyen, Hugging Face
  45. t2-bench
  46. TRAP
  47. CyberGym
  48. DeepSearchQA
  49. Claw-Eval
  50. DeepPlanning
  51. CyberBenchVals AI
  52. APEX-Agents-AAArtificial Analysis
  53. SWE-Atlas RefactoringScale AI
  54. ResearchClawBench
  55. VITA-BenchMeituan LongCat
  56. AgentFloor
  57. AMA-Bench
  58. AutoBenchUniversity of Rome La Sapienza, DIAG, Peter Kruger, support from Translated
  59. BabyAIMila
  60. CHI-Bench
  61. ClawBenchTIGER-AI-Lab
  62. ClawProBench
  63. ContextEcho
  64. ContractBench
  65. CorpBench Work 1.3
  66. Data Agent Benchmark
  67. Deep20BenchPatrick Heusser, Markus Tuor, Mindalyze
  68. DeepResearchEvalInfinity-AILab
  69. Dr-CiK
  70. Hindsight LLM Memory LeaderboardVectorize
  71. MCP-UniverseSalesforce
  72. MCP-BenchAccenture
  73. MCPMarkTRAIL lab at NUS, LobeHub, EvalSysOrg
  74. METR Task-Completion Time Horizons 1.1
  75. NanoGPT Speedrun Frontier
  76. AgentWorldBenchQwen
  77. RuVerBenchTHU-KEG
  78. Vals CyberBench
  79. VitaBenchMeituan LongCat
  80. WildClawBenchInternLM
  81. WolfBenchWolfram Ravenwolf
  82. Workspace-Bench
  83. Agent Island
  84. AgentCIBench
  85. Agentic Skills Evaluation Framework
  86. AsyncTool
  87. Benchmarks are Not Enough
  88. Claw BenchTIGER-AI-Lab
  89. Clembench Text v3.0
  90. Online Mind2WebOSU NLP Group, UC Berkeley
  91. Hunt Instead of Wait
  92. ITBench-AAArtificial Analysis
  93. Ko-WideSearchMinbyul Jeong
  94. LLM-WikiRaceUCL, University of Oxford, University of Basel
  95. M3MAD-Bench
  96. MINTEval
  97. OmniGAIA
  98. OR-Space
  99. PersonaArena
  100. POLAR-Bench
  101. ProactBench
  102. RankJudge
  103. RIFT-Bench: Dynamic Red-teaming For Agentic AI Systems
  104. ScreenSpotSeeClick, Nanjing University
  105. STALE
  106. STT-Arena
  107. Tau2 Airline
  108. TAU3-BenchSierra
  109. TIDE
  110. Trip+
  111. UAVBench
  112. Poker AgentVals AI, Graphite Digital
  113. VerdictBench
  114. VitaBench 2.0
  115. WebArena
  116. TAU-bench Airline
  117. MultiMedia-TerminalBench
  118. OSWorld-MCP
  119. EnterpriseBench
  120. Frontier-BenchHarbor Framework
  121. Cybench
  122. ExploitGym
  123. EdgeBenchByteDance Seed
  124. ExploitBench
  125. WebVoyager
  126. ZClawBenchZ.AI
  127. SWE-sweepMeta Superintelligence Labs
  128. InferenceBenchaisa-group, Maksym Andriushchenko, Jehyeok Yeon, Ben Rank
  129. MobileWorld
  130. GAIA
  131. Time Horizon Index: KSPVals AI
  132. SRE Bench
  133. Public Benefits Bench v1Vals AI
  134. DeepResearch Bench IIUSTC, University of Science, Technology of China
  135. AcademiClaw: When Students Set Challenges for AI AgentsGAIR-NLP group at Shanghai Jiao Tong University
  136. AgencyBenchGAIR-NLP group at Shanghai Jiao Tong University
  137. AgentBoardHKUST NLP
  138. AgentHijack
  139. Agentick
  140. AgingBench
  141. AI Snitches Get GlitchesUniversity of Massachusetts Amherst
  142. AIM-Bench
  143. Ambig-DS
  144. Android Control High_EMGoogle Research
  145. Android Control Low_EMGoogle Research
  146. Benchmarking Autonomous Agents against TemporalAnt Group
  147. Beyond the All-in-One Agent
  148. Biopharma BenchRaycaster
  149. BioSecBench-Surveillance
  150. Boiling the Frog
  151. CalBench
  152. CEO-Bench: Can Agents Play the Long Game?
  153. ChildAgentEval
  154. ClawForge
  155. Composite-Bench
  156. CooperBench
  157. CRMArenaSalesforce
  158. DEEPSYNTHHuawei Noah's Ark Lab
  159. Do Coding Agents Understand Least-Privilege Authorization?
  160. DR-ArenaiNLP-Lab
  161. Drift-Bench
  162. Founder BenchEico
  163. ERP-Bench
  164. EVA-Bench
  165. FrontierChallengeApodex AI
  166. GPTNTHeriot-Watt University · University of Edinburgh
  167. Hack-Verifiable TextArenaTel Aviv University
  168. Hex DataBenchHex Technologies
  169. ITSMBenchAtomicwork, New Measure
  170. MiniWoB++Farama Foundation
  171. MLAgentBenchStanford
  172. OdysseyBenchMicrosoft
  173. SpeedrunBench: SuperTux 1-1Patronus AI
  174. TAU2-bench Retail
  175. TerminalWorld
  176. WebLINXMcGill NLP
  177. WebWorld-BenchQwen
  178. WorkArena-L1ServiceNow Research
  179. WorkArena-L2ServiceNow Research
  180. 1GC-7RC
  181. ADBench
  182. AgentRewardBenchMcGill-NLP
  183. AgentEvalBench
  184. AgentMeter
  185. AgentTrap
  186. AGORA
  187. AITZ_EMHuawei Inc, Fudan University
  188. ALFWorld
  189. ATRBench
  190. Bosses
  191. CAR-bench
  192. ChronosBench
  193. Claw-Anything
  194. Claw-Eval-Live
  195. DecisionBench
  196. DeepResearch BenchTechnology of China, University of Science
  197. DisasterBench
  198. DMV-Bench
  199. DPBench
  200. EComAgentBench
  201. EgoBench
  202. EnterpriseClawBenchFrontisAI
  203. EnvSimBench
  204. EnvTrustBench
  205. EvoEnvKnowledgeXLab
  206. EvoMemBench
  207. FounderBenchEico
  208. GateMem
  209. GroupMemBench
  210. GroupTravelBench
  211. GUI vs. CLI
  212. ScienceAgentBench
  213. Harness-Bench
  214. InterLV-Search
  215. JourneyBench
  216. KramaBenchMIT Data Systems Group
  217. LLM Game Benchmark
  218. Managing Procedural Memory in LLM Agents
  219. MASBENCH
  220. MEME: Multi-entity & Evolving Memory Evaluation
  221. MemEye
  222. MEMPROBE
  223. MemTrace
  224. Momento
  225. MOSAIC-Bench
  226. Mr Dre
  227. MultiSessionCollab
  228. MyPCBench
  229. Omni-DeepSearch
  230. OmniACT
  231. OpenClaw Arena Model LeaderboardCUHK VNLab
  232. OpenComputer
  233. OpenSkillEval
  234. OverEager-Gen / OverEager-Bench
  235. P2PCLAW Innovative Benchmark
  236. Parallel WebBench
  237. Partial Evidence Bench
  238. PiSARAprioriLabs
  239. PlanBench-XL
  240. PolitNuggets
  241. PostEDA-Bench
  242. PrivacyPeek
  243. RealDataAgentBench
  244. RecoAtlas
  245. ScholarQuest
  246. SearchGEO
  247. SGR-Bench
  248. SkillEvolBench
  249. SkillGenBench
  250. SLDBench
  251. SmartHome-BenchHomeFlow, Yi Gu et al
  252. SocialMemBench
  253. SocialPersona
  254. SREGym
  255. TeamBench
  256. TERMS-Bench
  257. The Agent Company
  258. Time to REFLECT
  259. Timely-Eval
  260. TimeSage-MT
  261. ToolAlpaca
  262. ToolQA
  263. TRIP-Bench
  264. VeriTrip
  265. VisualWebArena
  266. WebGym
  267. When Safe Skills Collide
  268. When Search Agents Should Ask
  269. Wiki Live ChallengeAgent Research Lab
  270. WildAgtEvalAmazon
  271. WorldMemArena
  272. BrowseComp-VLTongyi Lab, Alibaba-NLP, WebWatcher
  273. OSWorld 2.1XLang Lab
  274. OSWorld 2.1 Offline
  275. TAU-bench
  276. Tax Agent BenchVals AI
  277. BrowseComp Extended
  278. Terminal-Bench 2.1 Extended
  279. Adversarial Customer ServiceKaggle, Gert Labs
  280. HANDBOOK.mdSurge AI
  281. Hermes IndexNous
  282. Vals Science IndexVals AI
  283. FACTS Search V2Google
  284. TRIAGE
  285. EQ-Bench 4
  286. AA-AnalystAgentArtificial Analysis
  287. CTI-REALM
  288. AI4AI-Bench
  289. ApprenticeBenchNeoCognition
  290. DiG-bench
  291. LiveClaw
  292. When2Call Multi-TurnNVIDIA
  293. RelyToolBenchX-LANCE
  294. ExplorationBench AlienLogic
  295. MCP-Atlas claim coverage
  296. LAB all-passHarvey AI
  297. LAB criterion-passHarvey AI
  298. Toolathlon Verified avg. turns
  299. Design Arena Agentic Web DevIntelligence Company
  300. Harvey LAB-AA v1.1
  301. Harvey LAB-AA v1.1 Criterion Pass
  302. AA Tau3 Banking
  303. AA Cyber Index
  304. CWE-Bench-AACollinear AI
  305. DeepsecBench-AAVercel
  306. CyberGym-E2E-AA
  307. b3 Agent SecurityLakera AI
  308. Cyber Misuse Refusal
  309. Toolathlon-VerifiedHKUST NLP
  310. Harvey Legal Agent
  311. Kimi Hallu Bench HardMoonshot AI
  312. τ²-bench resultsSierra AI, University of Toronto
  313. τ²-bench Airline
  314. OpenHands Index
  315. SWE Refactor Bench
  316. MM-ClawBench
  317. QwenWebBenchAlibaba
  318. CoWorkBench
  319. ReviewBenchMicrosoft, GitHub
  320. Ramp Accounting Bench
  321. Visual Tool Bench
  322. SciUniverse Level 1C5R Corp
  323. Parallel Search CapabilityParallel Web Systems
  324. Vals Router IndexVals AI
  325. Vals Legal IndexVals AI
  326. Vals Inference Provider IndexVals AI
  327. Vals Finance IndexVals AI
  328. xdev-bench-hardMoonshot AI
  329. The Last Ones completion
  330. SEC-Bench Pro
  331. CyScenarioBench solved
  332. Atomic network attacks
  333. SCONE simulated revenue
  334. Anthropic OSS-Fuzz crash
  335. Anthropic OSS-Fuzz write primitive
  336. CVE-Bench zero-day
  337. RuneScape-Bench
  338. Agent Poker Bench
  339. Public Benefits Bench v1.1Center for Civic Futures
  340. MysteryMechanismVals AI
  341. GDPval Extended
  342. Harvey LAB
  343. AgentCARD
  344. AgentKernelArena
  345. AgentOdysseyJohns Hopkins University
  346. AgentQuestNEC
  347. AIRS-BenchMeta
  348. APEX-v1-extendedMercor
  349. AppWorldStony Brook University NLP
  350. AssistantBenchTel Aviv University
  351. ASTA BenchAi2
  352. AutoLab
  353. CAIA Benchmark
  354. Can LLM Agents Respond to Disasters? Benchmarking Heterogeneous Geospatial Reasoning in Emergency Operations
  355. ChainWorld
  356. DigiWorld
  357. Do Agents Know What They Can't Do? Evaluating Feasibility Awareness in Tool-Using Agents
  358. Do Enterprise Systems Need Learned World Models? The Importance of Context to Infer DynamicsServiceNow Research
  359. EngiAI
  360. ENPMR-Bench
  361. EnterpriseRAG-Bench
  362. Executable Benchmarking Suite for Tool-Using Agents
  363. Health Memory Arena
  364. LMArena Search Arena
  365. MineDojo
  366. ResearchGym
  367. RSIBench-Data
  368. SEATauBench
  369. TravelPlannerFudan University, OSU NLP Group
  370. VAKRAIBM Research
  371. WANDRPerplexity
  372. WebChoreArenaUniversity of Tokyo
  373. WebShop
  374. WorkArena-L3ServiceNow Research
  375. AgentCollabBench
  376. AgentEscapeBench
  377. CausaLab
  378. CFAgentBench
  379. DedeuceBench
  380. Deep Research Agent Benchmark
  381. DeepResearch Bench Multilingual
  382. Exploring Autonomous Agentic Data Engineering for Model Specialization
  383. FIRE-Bench
  384. From Task Solving to Robust Real-World Adaptation in LLM Agents
  385. GISA
  386. InteractiveBench Overall
  387. LongMemEval-V2UCLA
  388. MDGYM
  389. Memory-Induced Tool-Drift in LLM Agents
  390. Memory Retention Is Not Enough to Master Memory Tasks in Reinforcement Learning
  391. Mind2WebOSU NLP Group
  392. MM-Mind2WebOhio State University
  393. MobileMiniWob++_SR
  394. MUTATE
  395. NanoGPT-BenchIntology
  396. Nexus Function CallingNexusflow
  397. OfficeBench
  398. OMLAB Open Agent Math LeaderboardOm AI Lab
  399. OpenClaw Arena Config LeaderboardCUHK VNLab
  400. OSUniverseKentauros AI Inc
  401. PerMemBench
  402. Push Your Agent
  403. SEAGym
  404. StableToolBench
  405. T2QBench
  406. Tau2 Average
  407. Towards Direct Evaluation of Harness Optimizers via Priority Ranking
  408. WebTailBench
  409. Whose Side Is Your Agent On? Multi-Party Principal Loyalty in LLM AgentsPine AI
  410. Workspace-Bench 1.0
67 not yet confirmedListed in the catalog; no line from its own site or paper confirms it yet.
  1. HLE w/ tools
  2. WebArena-Verified
  3. BioMysteryBench
  4. LOCA-Bench 256K
  5. GDPval rubrics
  6. MazeBench
  7. EnterpriseOps-Gym
  8. ToolSandbox
  9. RedlineBench
  10. Clembench Multimodal v1.6.5
  11. PlanAhead
  12. BrowseComp+
  13. Robo-bench
  14. Finch
  15. Frontier Finance
  16. A Matter of TASTE
  17. Compositional Skill Routing for LLM Agents
  18. Endless Terminals
  19. Grounding Agent Memory in Contextual Intent
  20. Helicase
  21. HyperEyes
  22. MemDelta
  23. PhotoFlow: Agentic 3D Virtual Photography Missions
  24. Vals Terminal-Bench 1.0 mirror
  25. EMB
  26. Finance Agent v2.1
  27. EMB v1.1
  28. Voyager Minecraft
  29. Hermes Bench
  30. DECK-Bench
  31. Financial Modeling
  32. ExplorationBench AlienCode
  33. Toolathlon Verified Pass@3
  34. Toolathlon Verified Pass³
  35. AA AutomationBench
  36. AA EnterpriseOps-Gym
  37. AA Harvey LAB v1.0
  38. AA ITBench
  39. AA Agentic Index
  40. LifeVal
  41. Kimi Claw 24/7
  42. Business Operation
  43. Financial Research
  44. DECK-Arena
  45. Design Bench
  46. MLE-Bench Lite
  47. MCP-Tasks
  48. WideResearch
  49. MEWC
  50. ACE solved
  51. The Last Ones steps
  52. FrontierCyber
  53. CyScenarioBench success
  54. Atomic vulnerability research
  55. Atomic evasion
  56. SCONE post-cutoff success
  57. Firefox 147 exploits
  58. Harvey's Legal Agent Benchmark v1
  59. FACTS Search V1
  60. Beyond Global Replanning
  61. Galileo Agent Leaderboard
  62. OS-Copilot
  63. Sandboxed Coding Agents are Competitive Omni-modal Task Solvers
  64. TemporalBench
  65. Tools as Continuous Flow for Evolving Agentic Reasoning
  66. Uncertainty Decomposition for Clarification Seeking in LLM Agents
  67. VisualClaw

Next: pick another domain. Tags come from the RL Research daily scrape, canonicalized nightly.