Domain on the map · as Coding

RL environments and benchmarks for private codebases

Coding work inside proprietary repositories rather than public open-source projects, to test how models handle unfamiliar company code they have never seen.

RL environments
7
AI benchmarks
24
Shares entries with
4 categories

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

The galaxy could not load. See every vendor on the board.

Coding, live from the universe. Hover a star for its name; click to open it. Open it in the universe

In short

What are RL environments for private codebases?
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 private codebases, that means coding work inside proprietary repositories rather than public open-source projects, to test how models handle unfamiliar company code they have never seen.
Who builds RL environments for private codebases?
Turing, Deeptune, Proximal, Mechanize, Huzzle Labs and Datacurve, among 7 confirmed below.
What are the main private-codebase coding benchmarks?
Terminal-Bench 3.0, CEO Bench, FrontierSWE (v1), CyberStrike, SciCode++ and Open-MM-RL, among 14 confirmed below.

RL environments 7

  1. Turing
  2. Deeptune
  3. Proximal
  4. Mechanize
  5. Huzzle Labs
  6. Datacurve
  7. Refresh

AI benchmarks 24

  1. Terminal-Bench 3.0Harbor Framework
  2. CEO Bench
  3. FrontierSWE (v1)
  4. CyberStrike
  5. SciCode++
  6. Open-MM-RL
  7. SDLCbench
  8. CompanyBench
  9. EKWbench
  10. Medical Bench
  11. OdysseysMega
  12. gba-eval-attempts and per-model gba-eval-attempt-* repos
  13. BlenderBench
  14. InsureBench: insurance underwriting and claims benchmark
10 not yet confirmedListed in the catalog; no line from its own site or paper confirms it yet.
  1. FrontierSWE v2
  2. Turing Terminal-Bench Data
  3. KernelQuest
  4. PLSQLBench
  5. Turing Frontier Agentic STEM
  6. Gauntlet 4K RLVR dataset, access by request
  7. OpenEnv Visual Memory and Spreadsheet environments on Hugging Face
  8. Robustly Evaluating Cyber Capabilities
  9. HuzzleWorld computer-use models
  10. computer-1 computer-use harness in Harbor

Data 32

  1. Abundantbased here
  2. Datacurvebased here
  3. Emulatedbased here
  4. General Reasoningbased here
  5. Habitat Incbased here
  6. Parsewavebased here
  7. pre.devbased here
  8. Revelobased here
  9. AfterQueryalso covers it
  10. Anthromindalso covers it
  11. Aviroalso covers it
  12. Collinearalso covers it
  13. Contra Labsalso covers it
  14. Cuaalso covers it
  15. Good Start Labsalso covers it
  16. Handshakealso covers it
  17. HUDalso covers it
  18. Huzzle Labsalso covers it
  19. Markovalso covers it
  20. Mercoralso covers it
  21. micro1also covers it
  22. Patronus AIalso covers it
  23. ReasonCorealso covers it
  24. Refreshalso covers it
  25. Surge AIalso covers it
  26. Symbalalso covers it
  27. Thetaalso covers it
  28. Tolokaalso covers it
  29. Trajectory Labsalso covers it
  30. Turingalso covers it
  31. Ulamalso covers it
  32. Vetto AIalso covers it

Domains that overlap with Private Codebases

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