Price cut, bill up 31%: new model's self-checking eats Anthropic's token discount
tengyanAI · x · 2026-11-11
AWS CEO Matt Garman says every 50% drop in inference prices brings 10x more usage, that inference is under 1% of its eventual scale, and that AWS is adding a GPU every 14 seconds with capacity shortages lasting years.
The author illustrates the flip side with Anthropic's Fable 5.2 launch: token prices fell 20% ($10/$50 to $8/$40 per million tokens), yet average cost per finished task on SemiAnalysis's GDPx benchmark rose 31% ($67.16 to $88.15). The new model writes hundreds or thousands of test cases per job (vs 20 before) and spins up subagents to recheck answers — which also took it to #1 on the benchmark. The discount got eaten by the model's diligence.
More from Infra
- 40 of 41 Data Center Ballot Measures in 2026 Midterms Went Against AI Builders — Acceptable-Offer-518 · 2026-11-12
- Azure Day at KubeCon: AKS Runs 22,655 Agents on 65 Nodes in Live Demo — davemccollough · 2026-11-12
- Gas-for-AI-datacenters forum co-located with CoRL shows energy demand spillover — carlosdponx · 2026-11-12
- a16z podcast: Mike Volpi on Cisco's 70+ acquisition playbook and the AI buildout — a16z · 2026-11-12
- AMD to pass Intel in server revenue by mid-2027 as share gains 6 points a year — BenBajarin · 2026-11-12
- Study: Reflective Data Center Roofs Actually Cool Daytime Ground by 1.8-2.2°C — SumitGup · 2026-11-12