Gemini 3.8 Flash Took 5 Minutes for a Grammar Check as 3.6/3.7 Face Deprecation
Party-Change3075 · reddit · 2026-10-07
- A Reddit user reports Gemini 3.6 and 3.7 Flash are now marked "Leaving Soon," while the replacement 3.8 Flash feels painfully slow.
- His test: a simple grammar-check on a short text took 3.8 Flash a full 5 minutes to generate (screenshots attached) — unacceptable for a tier branded "Flash."
- With Gemini 4.0 presumably on the horizon, he questions why Google is rushing to deprecate the only fast, reliable models, and asks whether others are seeing similar latency on 3.8.
More from Models
- Google researcher: Reflection's Beam is a "laundered" Chinese model — and that's actually safer — ibab · 2026-10-07
- Prime Agent demoed with new Qwen3.8-27B on Nebius: persistent Python env, self-refining harness — Arindam_1729 · 2026-10-07
- Mistral ships 1T-param Large 4, Google rolls out Nano Banana 2.1, OpenAI beta-launches Decisions API — testingcatalog · 2026-10-07
- ChatGPT hits 5.92B visits in September, its second-biggest month ever — gaganghotra_ · 2026-10-07
- Codex runs GPT-6 with far less reasoning than the API — up to 12x lower on max effort — RexDouglass · 2026-10-07
- Unconfirmed: Claude Opus 5.5 spotted rolling out on Grok bot — XFreeze · 2026-10-07