GPT-6 Astra Downgrade Was Real: OpenAI Acknowledged and Partially Fixed It
SteveEricJordan · reddit · 2026-09-12
A Reddit post argues that the GPT-6 Astra downgrade was real — and that OpenAI has acknowledged and partially fixed it — while calling out the recurring pattern of shipping a model, waiting for benchmarks and hype-driven subscriptions, then quietly dialing it back to save costs.
Key points:
- The author compares the practice to a cleaning service that starts cleaning half the apartment for the same price, arguing it should be illegal
- Proposes that benchmark sites regularly retest shipped models, since visible leaderboard drops would deter downgrades
- Notes OpenAI already paused the $200 subscription due to compute shortages amid Astra hype, making subtler rate-limiting plausible
- Concedes most downgrade posts are vibes and placebo, but says this case is verifiable and users should speak up
Related event: GPT-6 Astra Hit by Silent Downgrade Claims; OpenAI Admits and Fixes(10 posts)→
More from Models
- Open TTS Leaderboard adds 14 languages, new LFM3-Audio-2B and maya2-native models — realmrfakename · 2026-10-31
- Developers say the Linux Codex app is broken and neglected — eigenron · 2026-10-31
- Kimi now accepts Russian Mir cards directly: Pro subscription at $39/month without intermediaries — teortaxesTex · 2026-10-31
- Nat Lambert: Open models could be what holds the AI ecosystem together — natolambert · 2026-10-31
- Gemini 4 Carbon debuts #1 on APEX-Agents, the only model averaging above 90% — Swarooprm7 · 2026-10-31
- Sarvam quietly builds an India-focused opencode with 'local inference', gets mocked for the definition — kingslyj · 2026-10-31