Reddit 用户实测吐槽:OpenAI 编码额度半小时烧完,竞品能用一整天
Presentation_Past · reddit · 2026-11-21
A Reddit user subscribed to Claude, Google (Astra) and OpenAI (Codex) complains OpenAI can't even stay quota-competitive: in real coding projects, Opus — and even Neon, widely criticized for burning quota fast — consistently deliver more usable usage. One 5-hour Neon session completes complex multi-milestone projects, while Astra dies in half an hour having produced a fifth of the work. With OpenAI already behind Google on models, the author says he'd cancel if not for ChatGPT in daily chatting.
「模型」频道最新
- Workspace 付费用户被挡在 Gemini 门外:自定义域名账号无法使用 — film_girl · 2026-11-22
- Lean 社区验证 1700 万行代码,确认 Claude 参与证明黎曼猜想 — Polymarket · 2026-11-22
- Claude 预测:AI 五年内将攻克数学证明与电池新材料 — alex_verem · 2026-11-22
- 研究者质疑:OpenAI 中等推理档无法防蒸馏,防了也没用 — adam_karvonen · 2026-11-22
- 9B 模型跑出 59 tok/s:SenseNova-Loop-9B 在 16GB 显卡实测 — AIFlow_ML · 2026-11-22
- Anthropic 15亿美元版权和解未付款,被索赔每本书涨到2万 — TuhinChakr · 2026-11-22