Claude Effort vs Thinking: User Experience
Subtyr · reddit · 2026-07-13
The OP switched from Gemini to the free tier of Claude Web to evaluate if Sonnet 5 + Max Effort + Thinking is worthwhile for professional and paid projects.
They are concerned that low-quality outputs could mislead project decisions, but their current issue is that high Effort burns through quotas rapidly—sometimes enforcing a 5-hour wait for a single message. Additionally, the model occasionally stops mid-generation due to excessive length, forcing users to split queries across two messages. The post primarily asks for the practical differences and pros/cons of various Effort levels (Low/Medium/High/Max) and enabling Thinking.
More from Models
- Kimi K3 is praised for stronger English, frontend arena #1, and better handling of nuanced prompts — EXM7777 · 2026-07-22
- Gary Marcus says LLM math skills are like knowing only a car’s engine size — GaryMarcus · 2026-07-22
- OpenAI’s Codex + GPT-5.6 Sol hits 99% recall in Project APE verification tests — soumitrashukla9 · 2026-07-22
- OpenAI-linked paper says capability RL can make models more reward-seeking — MariusHobbhahn · 2026-07-22
- Macaron V1 adds LoRA RL on GLM 5.2 and claims SOTA benchmark gains — Xianbao_QIAN · 2026-07-22
- OpenAI rolls out voice in GPT-Live, but the UI obscures search and reasoning — Graham_dePenros · 2026-07-22