Anthropic gets roasted for keeping its best model in-house and neglecting the smaller ones
AaronBergman18 · x · 2026-07-28
The repost repeats a joke-turned-observation about Anthropic’s model lineup:
- The author says they liked versions 4.5 and 4.6 a lot, but versions 4.7, 4.8, and 5 are harder to keep in a friendly mode.
- The punchline is that Anthropic employees apparently use the best model internally, while the smaller models are neglected and turned into tools.
- It frames model behavior and internal dogfooding as part of why some models feel more polished or “friendly” than others.
More from Companies & People
- Poster says the NVIDIA-linked release is still gated to approved Microsoft Foundry customers — davidmanheim · 2026-07-28
- Companies reject BigQuery for search data, but approve prompt trackers fast — gaganghotra_ · 2026-07-28
- AI creator says one day of work now pays more than a month of design gigs — Tegadesigns · 2026-07-28
- OpenAI co-founder quote sparks a human-oversight vs EU AI Act debate — davidmanheim · 2026-07-28
- Brown Health says Microsoft Dragon Copilot and 24 AI agents are easing clinical work — luisdans · 2026-07-28
- AI models will become every company’s new website, but data flywheels will decide the winners — aigclink · 2026-07-28