Directory label

Tool-use benchmarks for LLMs and tool-use 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.

Calling external tools and APIs correctly: choosing the right function, passing valid arguments and using the results to finish a task.

RL environments
3
AI benchmarks
26
Shares entries with
3 categories

Snapshot 2026-11-12. How categories are built · Where the data comes from

In short

What are RL environments for tool use?
A simulated workplace where an AI agent practises a task (a code repository, a browser, business software) and an automatic grader scores each attempt. Labs train on thousands of these attempts. Bought to improve models. For tool use, that means calling external tools and APIs correctly: choosing the right function, passing valid arguments and using the results to finish a task.
Who builds RL environments for tool use?
Chakra Labs, Pareto and Markov.
What are the main tool-use benchmarks?
BFCL-V4, BFCL_v3_MultiTurn, Leap, Attune, TRACE and ComplexMCP, among 22 confirmed below.

RL environments 3

  1. Chakra Labs
  2. Pareto
  3. Markov

AI benchmarks 26

  1. BFCL-V4UC Berkeley
  2. BFCL_v3_MultiTurnUC Berkeley
  3. LeapPareto
  4. AttuneThoughtful Lab
  5. TRACE
  6. ComplexMCP
  7. ACEBench
  8. ComplexFuncBench
  9. TOBench
  10. Trust No Tool
  11. When Simulation Lies
  12. BFCL v2UC Berkeley
  13. API-Bank
  14. AWARE-USMehmet Kurmaz
  15. Beyond Function Calling
  16. Gorilla Benchmark API BenchUC Berkeley
  17. MANTRA
  18. MAVEN-Bench
  19. TIDE-Bench
  20. Time Puzzles
  21. ToolPrivacyBench
  22. LLM Agents Already Know When to Call Tools -- Even Without ReasoningTrustworthy ML Lab
4 not yet confirmedListed in the catalog; no line from its own site or paper confirms it yet.
  1. CUA data
  2. To Call or Not to Call
  3. TSCG
  4. Beyond Static Tools

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