The Rise of Generative AI Large Language Models (LLMs) like ChatGPT — Information is Beautiful Skip to content

The Rise of Generative AI Large Language Models (LLMs) like ChatGPT

Loading

A visualisation of major large-language models (LLMs), ranked by performance, using MMLU (Massive Multitasks Language Understanding) a benchmark for evaluating the capabilities of large language models.<br>" main source: Life Architect data<br>" see our data<br>" chart rendered with VizSweet<br>notes<br>The MMLU rating consists of 16,000 multiple-choice questions across 57 academic subject (link)<br>MMLU has some critiques, mainly that LLM creators may be wise to the metric and be pre-training their models to answer MMLU questions. (New York Times article). And here’s a general cautionary article.<br>If you’re interested in ongoing metrics , LLMArena has a leaderboard ranked subjectively by tens of thousands of users. Note: it privileges English language LLMs (there are no Chinese models on there for example). Intel also run a leaderboard with many measurements.<br>Note: we excluded specialised maths and coding LLMs like DeepSeek Coder, Mathstral, Granite Code etc. Let us know if we’ve missed any LLMs.

A visualisation of selected lawsuits from over 100 filed against AI companies as of June 2026 .<br>" See the data & research<br>SOURCES: Wired, ChatGPT is Eating the World, news reports<br>Learn to Create Graphics Like This<br>Most are copyright claims by authors, record labels, newsrooms, academic publishers and um, Disney who say major AI models like OpenAI’s ChatGPT and Anthropic’s Claude have been trained illegally on their creative and commercial output.<br>These are the opening salvos in a structural conflict over who owns the raw material of the internet. And what constitutes “fair use” of that material in the age of AI. And whether there will be push back agains the inherently extractive nature of AI models.<br>Most cases haven’t been decided yet. But some are worthy of detailing:<br>Bartz vs Anthropic<br>Three authors – Andrea Bartz, Charles Graeber, and Kirk Wallace Johnson – sued Anthropic, alleging their flagship AI model Claude was trained on millions of pirated e-books downloaded from online “shadow libraries" without permission.<br>Anthropic settled the case in April 2026 for $1.5billion – the largest copyright settlement in US history.<br>How much individual authors will receive in compensation is currently unknown, especially now the case is delayed while lawyers’ excessive fees ($320m!) and other legal tangles are ironed out.<br>You can see if your work or favourite book are included in the training set.<br>Pile On<br>Similar cases like Carreyrou vs Open AI , Kwon vs Anthropic , Cruz vs Anthropic , Hobbs vs Meta are using the same tack, accusing the tech firms of torrenting millions of copyright books from shadow websites rather than lawfully licensing them. Given the Bartz vs Anthropic precedent, these cases may have a strong chance of succeeding.<br>Simultaneously, Elsevier, a large scientific publisher, alongside four other major publishers — Hachette, Macmillan, McGraw Hill, and Cengage — and the best-selling novelist and lawyer Scott Turow are directly accusing Meta and Mark Zuckerberg of doing the same thing for their Llama model.<br>Disney<br>Disney is going after popular and powerful image generator MidJourney, calling it a “bottomless pit of plagiarism” for its alleged reproductions of the studios’ best-known characters.<br>Music Models<br>At the same time major record labels like Universal Music Group (UMG) and musicians rights holders like Danish Koda are suing the dominant music AI model Suno for apparently training itself on all of Western music. “We are witnessing the largest music theft in history,” Koda’s announcement of the lawsuit reads.<br>resources & further detail<br>" All cases tracked @ ChatGPT is Eating The World<br>" Every Copyright Court Case Visualised (Wired)<br>" AI Watchdog from The Atlantic<br>" Another (more legalese) tracker<br>total lawsuits<br>(as of Jun 2026)<br>Open AI – 24<br>Meta – 14<br>Microsoft – 10<br>Anthropic – 9<br>NVIDIA – 7<br>Perplexity – 7<br>Apple – 5<br>Google – 5<br>RunwayAI – 4<br>Adobe – 4<br>Suno – 4<br>Stability – 3<br>Uncharted Labs – 3<br>Midjourney – 2

CHANGE LOG UPDATES<br>: 16th Jun – added Who’s Suing Whom graphic<br>: 4th Feb 2026 – added Kimi K2 Thinking, GPT 5.1 & 5.2, Gemini 3, Claude Opus 4.5, Mistral Larger 3, Nova 2 Pro, Kimi K2.5, DeepSeek v3.2, Grok 4.1<br>: 17th Sep – added ‘What are people using ChatGPT for?’<br>: 17th Sep – added SWE-Bench vs LLMArena score chart for major models<br>: 16th Sep 2025 – major update, adding Claude 3.7 & 4.0 sonnet, Claude Opus 4, Deepseek v3.1, Exaone 4.0, Exaone Deep, Gemini 2.0 Pro, Gemini 2.5 Pro, Gemini 2.5 Flash preview, Gemma 3, GPT 4.5, GPT-5, Grok 3 & 4, Hunyuan Turbo S, Hunyuan A13B, Llama 4 Maverick, Kimi K2, Mistral Saba, Mistral Medium 3, Mistral Small 3.1, OpenAI o3 & o4-mini, Pangu Ultra, Pangu Ultra MoE, Qwen 2.5-max, Qwen 3-235B-A22b, Qwen3-Next, Qwen3-Thinking-2507<br>: 20th Dec 2024 – added new ranking chart, added LMArena Scores<br>: 12th...