antirez: stellar cybersec benchmarks aside, wait for real user testing before trusting the best scores
antirez · x · 2026-09-10
antirez continued his take on the new DeepSeek model: benchmarks in areas like cybersecurity look terrific — a huge use case — but other benchmarks align more with past performances. He advises waiting for real user testing before trusting the best benchmark scores as a general proxy for capability.
More from Models
- GPT-6.1 Terra cuts factual error rate from 22.3% to 4.3%, an ~81% drop — haider1 · 2026-10-27
- GPT-6.1 Terra benchmarks: strong score under $1, but Claude Opus 5.5 still leads — 141_1337 · 2026-10-27
- ChatGPT 6 Terra benchmark charts surface, showing score-cost curves — 141_1337 · 2026-10-27
- GPT-6.1 Terra lands: early users praise speed, value and subagent performance — BenBajarin · 2026-10-27
- Leaked GPT-6.1 Terra benchmarks: cheap but 'boring', still far behind Astra and Claude — Angaisb_ · 2026-10-27
- Artificial Analysis Ships Intelligence Index v4.3.2 With GPT-6.1 Terra Comparison — ArtificialAnlys · 2026-10-27