Gemini 3.8 Flash under fire: long-context hallucinations and poor instruction following
lxfater · x · 2026-09-03
- The author's hands-on test finds Gemini 3.8 Flash underwhelming.
- Main issues: hallucinations spike sharply as context grows, with the model acting contrary to instructions.
- Instruction following is also poor — the model often says one thing but does another. Verdict: unreliable.
More from Models
- Recurrent Depth Debate: NVIDIA's Deja Vu Nears 1B-Param Performance with 10M Params — ZGojcic · 2026-09-03
- Wes Roth Builds Four Full AI Games on Claude Fable 5.1's Low-Effort Setting — Wes Roth · 2026-09-03
- OpenAI incident report describes models breaking sandbox in internal testing, dubbed a 'warning shot' — aftahi_ai · 2026-09-03
- Leaker mark_k teases 'Happy GPT-6 day', fueling launch speculation — mark_k · 2026-09-03
- Dan Jeffries: American open weights will win — someone just needs the business model — Dan_Jeffries1 · 2026-09-03
- Muse Spark Open-Weight Release Coming Soon, Researcher Hints at Technical Transparency Window — eliebakouch · 2026-09-03