Anthropic gave external researchers privacy-preserving access to real Claude usage for the first time. Stanford SALT, Oxford HIP Lab, and METR independently analyzed about 250,000 Claude.ai or Claude Code conversations from April–May 2026, with METR estimating real-world productivity gains from coding agents.

Key Takeaways

  • ✓Three external labs analyzed ~250,000 Claude.ai / Claude Code chats; SALT found over half involved consequential work that affects others or is hard to undo
  • ✓METR is estimating coding-agent productivity from Claude Code traces; early results show newer models save more time, and Claude's time estimates correlate with real developer times
  • ✓Aggregate data is public on Hugging Face (Anthropic/enabling-independent-research), with an open call for the next research cohort
⚡

Key Decision Metrics at a Glance

TL;DR Verdict
🏆 Benchmark RankRank
See how it ranks on the leaderboard
Full Leaderboard→
💰 Cost & PlansCompare
in: $5 / out: $25
Compare Plans↗
🛠️ Ecosystem & IDERules
Claude Code (Anthropic)
Setup IDE Rules↗
🧭

Heavy Claude Code use: compare subscription limits and API bills

Compare 40 dev plans & simulate token costs vs $20/mo subscriptions

⚔️ Compare Model→💰 Compare plans & limits↗
Action HubReady to adopt this in production?

Benchmark side-by-side against alternatives, or calculate monthly token cost vs subscription break-even.

⚔️ Launch VS Compare→💳 View Plans↗
Up NextClaudeDevs: Fable 5.1 live in Claude Code; 55.8% on Terminal-Bench 4.0
Read Now→