\n<\/p>\n
AI models are being cranked out at a dizzying pace, by everyone from Big Tech companies like Google to startups like OpenAI and Anthropic. Keeping track of the latest ones can be overwhelming.\u00a0<\/p>\n
Adding to the confusion is that AI models are often promoted based on industry benchmarks. But these technical metrics often reveal little<\/a> about how real people and companies actually use them.\u00a0<\/p>\n To cut through the noise, TechCrunch has compiled an overview of the most advanced AI models released since 2024, with details on how to use them and what they\u2019re best for. We\u2019ll keep this list updated with the latest launches, too.<\/p>\n There are literally over a million AI models out there: Hugging Face, for example, hosts over 1.4 million<\/a>. So this list might miss some models that perform better, in one way or another.\u00a0<\/p>\n OpenAI calls Orion their largest model to date<\/a>, touting its strong \u201cworld knowledge\u201d and \u201cemotional intelligence.\u201d However, it underperforms on certain benchmarks compared to newer reasoning models. Orion is available to subscribers of OpenAI\u2019s $200 a month plan.<\/p>\n Anthropic says this is the industry\u2019s first \u2018hybrid\u2019 reasoning model<\/a>, because it can both fire off quick answers and really think things through when needed. It also gives users control over how long the model can think for, per Anthropic<\/a>. Sonnet 3.7 is available to all Claude users, but heavier users will need a $20 a month Pro plan. <\/p>\n Grok 3 is the latest flagship model<\/a> from Elon Musk-founded startup xAI. It\u2019s claimed to<\/a> outperform other leading models on math, science, and coding. The model requires X Premium (which is $50 a month.) After one study found<\/a> Grok 2 leaned left, Musk pledged<\/a> to shift Grok more \u201cpolitically neutral\u201d but it\u2019s not yet clear if that\u2019s been achieved.<\/p>\n This is OpenAI\u2019s latest reasoning model<\/a> and is optimized for STEM-related tasks like coding, math, and science. It\u2019s not OpenAI\u2019s most powerful<\/a> model but because it\u2019s smaller, the company says<\/a> it\u2019s significantly lower cost. It is available for free but requires a subscription for heavy users.<\/p>\n OpenAI\u2019s Deep Research is designed for doing in-depth research<\/a> on a topic with clear citations. This service is only available with ChatGPT\u2019s $200 per month Pro subscription<\/a>. OpenAI recommends it<\/a> for everything from science to shopping research, but beware that hallucinations remain a problem<\/a> for AI.<\/p>\n Mistral has launched app versions of Le Chat<\/a>, a multimodal AI personal assistant. Mistral claims<\/a> Le Chat responds faster than any other chatbot. It also has a paid version with up-to-date journalism<\/a> from the AFP. Tests from Le Monde<\/a> found Le Chat\u2019s performance impressive, although it made more errors than ChatGPT.<\/p>\n OpenAI\u2019s Operator is meant to be<\/a> a personal intern that can do things independently, like help you buy groceries. It requires a $200 a month ChatGPT Pro subscription. AI agents hold a lot of promise, but they\u2019re still experimental: a Washington Post reviewer says Operator<\/a> decided on its own to order a dozen eggs for $31, paid with the reviewer\u2019s credit card.<\/p>\n Google Gemini\u2019s much-awaited flagship model<\/a> says it excels at coding and understanding general knowledge. It also has a super-long context window of 2 million tokens, helping users who need to quickly process massive chunks of text. The service requires (at minimum) a Google One AI Premium subscription of $19.99 a month.<\/p>\n This Chinese AI model<\/a> took Silicon Valley by storm<\/a>. DeepSeek\u2019s R1 performs well on coding and math, while its open source nature means anyone can run it locally. Plus, it\u2019s free. However, R1 integrates<\/a> Chinese government censorship and faces rising bans<\/a> for potentially sending user data back to China.<\/p>\n Deep Research summarizes Google\u2019s search results<\/a> in a simple and well-cited document. The service<\/a> is helpful for students and anyone else who needs a quick research summary. However, its quality isn\u2019t nearly as good as an actual peer-reviewed paper. Deep Research requires a $19.99 Google One AI Premium subscription.<\/p>\n This is the newest and most advanced version<\/a> of Meta\u2019s open source Llama AI models. Meta has touted this version<\/a> as its cheapest and most efficient yet, especially for math, general knowledge, and instruction following. It is free and open source.<\/p>\n Sora is a model that creates realistic videos<\/a> based on text. While it can generate entire scenes rather than just clips, OpenAI admits<\/a> that it often generates \u201cunrealistic physics.\u201d It\u2019s currently only available on paid versions of ChatGPT, starting with Plus, which is $20 a month.\u00a0<\/p>\n This model is one of the few to rival<\/a> OpenAI\u2019s o1 on certain industry benchmarks, excelling in math and coding. Ironically for a \u201creasoning model,\u201d it has \u201croom for improvement in common sense reasoning,\u201d Alibaba says<\/a>. It also incorporates Chinese government censorship, TechCrunch testing shows<\/a>. It\u2019s free and open source.<\/p>\n Claude\u2019s Computer Use is meant to take control of your computer<\/a> to complete tasks like coding or booking a plane ticket, making it a predecessor of OpenAI\u2019s Operator. Computer use, however, remains in beta<\/a>. Pricing is via API: $0.80 per million tokens of input and $4 per million tokens of output.<\/p>\n Elon Musk\u2019s AI company, x.AI, has launched an enhanced version of its flagship Grok 2 chatbot<\/a> it claims<\/a> is \u201cthree times faster.\u201d Free users are limited to 10 questions every two hours on Grok, while subscribers to X\u2019s Premium and Premium+ plans enjoy higher usage limits. x.AI also launched an image generator, Aurora, that produces highly photorealistic images<\/a>, including some graphic or violent content.<\/p>\n OpenAI\u2019s o1 family<\/a> is meant to produce better answers by \u201cthinking\u201d through responses through a hidden reasoning feature<\/a>. The model excels at coding, math, and safety, OpenAI claims<\/a>, but has issues deceiving humans, too.<\/a> Using o1 requires subscribing to ChatGPT Plus, which is $20 a month.<\/p>\n Claude Sonnet 3.5 is a model Anthropic claims as being best in class<\/a>. It\u2019s become known for its coding capabilities and is considered a tech insider\u2019s chatbot of choice.<\/a> <\/strong>The model can be accessed for free on Claude although heavy users will need a $20 monthly Pro subscription. While it can understand images, it can\u2019t generate them.<\/p>\n OpenAI has touted GPT 4o-mini<\/a> as its most affordable and fastest model yet thanks to its small size. It\u2019s meant <\/a>to enable a broad range of tasks like powering customer service chatbots. The model is available on ChatGPT\u2019s free tier. It\u2019s better suited for high-volume simple tasks compared to more complex ones.<\/p>\n Cohere\u2019s Command R+ model<\/a> excels at complex Retrieval-Augmented Generation (or RAG) applications for enterprises. That means it can find and cite specific pieces of information really well. (The inventor of RAG actually works at Cohere<\/a>.) Still, RAG doesn\u2019t fully solve AI\u2019s hallucination problem.<\/a><\/p>\n<\/div>\nAI models released in 2025<\/h2>\n
OpenAI\u2019s GPT 4.5 \u2018Orion\u2019<\/h4>\n
Claude Sonnet 3.7<\/h4>\n
xAI\u2019s Grok 3<\/h4>\n
OpenAI o3-mini<\/h4>\n
OpenAI Deep Research<\/h4>\n
Mistral Le Chat<\/h4>\n
OpenAI Operator<\/h4>\n
Google Gemini 2.0 Pro Experimental<\/h4>\n
AI models released in 2024<\/h2>\n
DeepSeek R1<\/h4>\n
Gemini Deep Research<\/h4>\n
Meta Llama 3.3 70B<\/h4>\n
OpenAI Sora<\/h4>\n
Alibaba Qwen QwQ-32B-Preview<\/h4>\n
Anthropic\u2019s Computer Use<\/h4>\n
x.AI\u2019s Grok 2\u00a0<\/h4>\n
OpenAI o1<\/h4>\n
Anthropic\u2019s Claude Sonnet 3.5\u00a0<\/h4>\n
OpenAI GPT 4o-mini<\/h4>\n
Cohere Command R+<\/h4>\n