Dario on open-weights risk; same model costs 37x more on WorkBuddy vs Doubao
kevinnbass · x · 2026-11-18
- The post relays Dario Amodei's argument that open weights escape vendor control, while closed deployment enables usage monitoring, access revocation, and behavior changes; the author frames this as monopoly ambitions cloaked in safety talk, noting Altman has voiced similar views.
- A quoted benchmark run: same models, same prompts, 34 runs each on Doubao Work vs WorkBuddy — the biggest gap was 37x. Kimi K3.1 Pro analyzing the same order sheet cost 0.13 RMB on Doubao Work vs 4.65 RMB on WorkBuddy, both scoring full marks.
- The core driver is credit pricing: 0.0076 RMB/point on Doubao Work vs 0.0175 RMB/credit on WorkBuddy, a 2x+ unit-price gap; with a DeepSeek V4.1 Flash limited-time discount, 5 Doubao tasks were free. Doubao was cheaper in 6 of 9 comparable pairs, though WorkBuddy won on GLM 5.3 order analysis and Kimi K3 PPT tasks on both price and quality score.
- Takeaway: identical models don't mean identical costs — platform credit pricing drives the bill; run light frequent tasks on Doubao Work, and dual-run important research/PPTs.
More from AGI Musings
- Are open-source models just linear combinations of Claude and GPT mind styles? — Zergylord · 2026-11-18
- Bank CEOs and NSA tell Congress AI-found vulnerability patches are behind nationwide outages — peterwildeford · 2026-11-18
- RSI Paper Sparks Debate: omarsar0 Shares L1-L5 Autonomy Scale to Define Recursive Self-Improvement — omarsar0 · 2026-11-18
- HUMAN.md: A Working Manifesto on What Humans Still Bring When AI Averages Everything — Dr_Atoosa · 2026-11-18
- Reddit asks: Hinton's predictions from three years ago still hold — what do big labs think? — SuburbanDad18 · 2026-11-18
- 300 Researchers Convene on AI Consciousness in Berkeley as AP and Vatican Declare AI Non-Sentient — ShakeelHashim · 2026-11-18