Claude 5.1 Costs Rise 20% Despite Cache Cuts Due to Bloat and Watermarks
op7418 · x · 2026-09-02
Anthropic released Fable 5.1 and Mythos 5.1 with a 75% cache price cut, but real-world tests by Artificial Analysis show a 20% cost increase for tasks due to 1.7x higher token output. Users report rapid quota depletion and critical issues: lack of forced tool choice, broken model switching (legacy models can't read thinking blocks), and intrusive invisible text watermarks degrading output quality.
Related event: Fable 5.1 Tops Benchmarks but Surging Costs Spark Debate(24 posts)→
More from Infra
- YC Paper Club Call: Optical Compute, Diamond Chips, and Bio-GPUs — ycombinator · 2026-09-02
- Texas Freezes Data Center Grid Connections Over Uncertain AI Power Demand — VraserX · 2026-09-02
- Google Infra Head: A Gemini prompt consumes energy equal to 7 seconds of TV — BenBajarin · 2026-09-02
- Running 104GB Qwen3.8-Flash-Next on 48GB Mac at ~12 tok/s — yogthos · 2026-09-02
- Not Diamond releases model routing method, cuts costs 20-80% — rohanpaul_ai · 2026-09-02
- Trading VRAM for FLOPs: Local Model Quantization Debate — max_paperclips · 2026-09-02