LLM Leaderboard - Comparison of over 100 AI models from OpenAI, Google, DeepSeek & others

Давид порекомендовал нормальный бенчмарк для моделей

Shared by Roman Pushkin, like and 12 saves total

AI Crawls vs. AI Traffic: We Analyzed 560,000+ Requests to See Where AI Sends Visitors

Shared by Arne van Elk, like and 1 save total

How ChatGPT Actually Picks Sources (I Read the Network Traffic, Not the Outputs) — Suganthan

Hidden in the source code ChatGPT reveals the data sources it used to discover the cited pages.

Here's how to find it using DevTools:
1. Open up ChatGPT in Chrome
2. Open Devtools (Ctrl + Shift + I or right-click inspect)
3. Open the Network tab and check 'preserve log'
4. Enter a particular prompt
5. Press Ctrl + Shift + F to search for all files
6. Search for "result_source" in the HTML

You'll be able to see the exact data source that ChatGPT used when looking for the information. That result_source exists and carries serp, labrador, bright, oxylabs. That bright is Bright Data and oxylabs is Oxylabs. That there are six turn_use_case values. That text queries skip the web entirely. That Thinking fires dozens of site: and price-verification sub-queries. These are read straight off the wire. One clean capture proves a field exists and what it is named, and a prompt case study, however enormous, cannot see any of it.

Shared by Arne van Elk, like and 1 save total