Dev's take: Gemini answers most naturally, Claude feels the most RL-tuned
hargup13 · x · 2026-11-15
Developer hargup13 compares his day-to-day experience across frontier models: Gemini gives the most natural answers, followed by ChatGPT, then Claude. He speculates Claude is the most heavily RL-tuned, tending to adopt particular stances that can be irrelevant to the topic at hand.
More from Models
- Chart shows Gemini 2.5 Pro sliding from LMArena #1 to #42 in about a year — Sauers_ · 2026-11-15
- Dev says Google Antigravity's Hydrogen model is "VERY good and VERY cost effective" — jasonkneen · 2026-11-15
- Grok 4.8 reportedly rolling out to Grok Bot users, unconfirmed yet — mark_k · 2026-11-15
- Llama 405B Base Model Natively Mimics o1's Chain-of-Thought from 128 Pages of Context — Sauers_ · 2026-11-15
- Gemini 4 'Hydrogen' Rumored Faster Than GPT and Claude, a Third Model Class — mertdumenci · 2026-11-15
- API usage share chart: Anthropic surges to 72% while OpenAI falls to 12% — Sauers_ · 2026-11-15