Tyler.is ≡ (@tylerwillis) 's Twitter Profile
Tyler.is ≡

@tylerwillis

Founder of Unsupervised.com (AI Data Analyst, insights found worth over $1B, raised $60M). I also invest in startups (11 Unicorns).

ID: 3916461

linkhttp://tyler.is calendar_today09-04-2007 15:49:08

564 Tweet

8,8K Followers

7,7K Following

Tyler.is ≡ (@tylerwillis) 's Twitter Profile Photo

The gap between open source models and proprietary ones has closed to near zero on benchmarks. I don’t know of anyone (beyond LM Arena) who’s done large scale user evals, but those ELO scores tell a similar story: best model is 1463; best open-source model is 1426 (breaking 1400

The gap between open source models and proprietary ones has closed to near zero on benchmarks.

I don’t know of anyone (beyond LM Arena) who’s done large scale user evals, but those ELO scores tell a similar story: best model is 1463; best open-source model is 1426 (breaking 1400
Tyler.is ≡ (@tylerwillis) 's Twitter Profile Photo

Anecdotally, LLM responses started feeling slow as GPT-5 got recommended more on my chats. GPT-5 response times were slow & Sonar has always been a bit slow. Haven’t run a benchmark yet, but feels like both have gotten much faster over the past ~36 hours. (Note: this is for