Insider: Anthropic's acceleration was mostly catchup — Claude 5.5 was the real breakout
kuza55 · x · 2026-11-20
Responding to the puzzle that Anthropic and OpenAI have been roughly tied for a year yet Anthropic is said to be 'much more accelerated,' kuza55 argues Anthropic was actually behind on model quality for most of its existence, and its recent acceleration is mostly catchup — though Claude 5.5 was a genuine breakout.
A representative AI-circle debate over how to read the relative pace of the two frontier labs.
More from Models
- Google brings Gemini 3.5 Transcribe voice editing to the Gemini API — GeminiApp · 2026-11-21
- Have frontier models found an exploit in LEAN? Reddit weighs testable signs — wabawanga · 2026-11-21
- Kapitsa runs Hy4-flash on M5 Max at 36 tok/s decode, 971 tok/s prefill at 32K — m4rkGS · 2026-11-21
- Learning to code now requires a $20/month AI subscription, Spanish devs warn of rising barrier — feregri_no · 2026-11-21
- Open-source AI notes: Inkling 2 Small tops Agent Arena, Alibaba to open-source Qwen4 small models — 0xsachi · 2026-11-21
- Prankster feeds Gemini Neon fake Riemann proof news and it completely takes the bait — corbtt · 2026-11-21