Opus 5.5's new Max effort mode burns 6x usage, sparking benchmaxxing claims
TigerConsistent · reddit · 2026-09-23
A Reddit user questions Anthropic's new Max effort mode on Opus 5.5, which consumes 6x the normal usage. If Anthropic benchmarks on Max while subscribers stick to Medium/High due to rate limits, users may never see the performance shown in benchmarks. The author calls the 6x usage overhead insane and wonders what Max is actually doing — a pointed challenge to the benchmark-vs-reality gap.
More from Models
- LIBERO experiments show stronger reasoning in GPT-6 variants means better robot manipulation — YuXiang_IRVL · 2026-09-23
- Diffusion Gemma is undervalued: real System-1 with both intelligence and speed — bingxu_ · 2026-09-23
- Claude Opus 5.5 models individual hair strands in Blender demo — majidmanzarpour · 2026-09-23
- Opus 5.5 on Perplexity: generous credits plus automatic prompt-injection checks for skills — ChrisUniverse · 2026-09-23
- Miles Brundage jokes new models will improve on "YouTube Poop bench" — Miles_Brundage · 2026-09-23
- Robot arm tests rank GPT-6 variants: better reasoning means better manipulation at 30x the cost — YuXiang_IRVL · 2026-09-23