'I miss 5.5': users say new models are better but lost their personality
Dapper_Ambition5963 · reddit · 2026-10-19
A legal professional recounts how Claude 5.5 helped him build a local LLM setup, explore AI governance as a career path, and even play Valheim together—arguing the newer Astra shows clear improvements but lacks 5.5's personality. The post highlights how model personality is becoming a real driver of user attachment.
More from Models
- Qwen3.8 27B benchmarks on NVIDIA RTX Spark: 19.3 tok/s decode, 100% accuracy — draginol · 2026-10-19
- Allie Miller: Claude leads on models, Codex on features — she uses both — alliekmiller · 2026-10-19
- Gemma 4 12b on NVIDIA RTX Spark laptop: 31.6 tok/s decode, 100% accuracy — draginol · 2026-10-19
- Same prompt, three designs: Claude vs ChatGPT vs Figma Make UI face-off — Tegadesigns · 2026-10-19
- Mac Studio M5 Ultra beats dual DGX Sparks in local LLM decode speeds — SumitGup · 2026-10-19
- Three mystery OpenAI models (Aurora red/green/alpha) spotted; likely image models, GPT Image refresh imminent — kimmonismus · 2026-10-19