56 AI practitioners vote: Terminal-Bench and TB Science top frontier-detection benchmarks
BenBlaiszik · x · 2026-11-16
Chris Barber asked 56 people in AI which benchmarks they'd trust most to indicate whether an LLM is frontier-level. Results:
- Terminal-Bench 4: 25 votes, #1
- Terminal-Bench Science: 20 votes, #2
- FrontierCode: 13 votes
- FrontierSWE: 10 votes
- Humanity's Last Exam: 8 votes
- Vending-Bench: 7 votes
- AutomationBench, DeepSWE: 5 votes each
Terminal/real-world task benchmarks clearly outrank exam-style ones like HLE, with agentic and SWE-focused benchmarks dominating the top of the list.
More from Models
- Liquid AI d1-2B decision model runs on-device: ~68ms gesture reading, 14 decisions/s, zero tokens — AccBalanced · 2026-11-16
- Anthropic user blasts its 'queuing steering' system as awful — willdepue · 2026-11-16
- Rumor: Google Prepping Strongest Gemini Model Yet, Codenamed Krypton — aziz4ai · 2026-11-16
- Blender watch test: DeepSeek does it for $0.80 vs Claude Fable's $119 — HankYeomans · 2026-11-16
- Users say Gemini turned colder after Nov 12 update, want old version back — Sean8277 · 2026-11-16
- Claude Code auto-memory injects context into every tool call; dev says disable it — nateberkopec · 2026-11-16