Keeping Up with the Latest AI Models

The pace at which AI models are being developed is staggering, with companies like Google, OpenAI, and Anthropic, as well as numerous startups, releasing new models constantly. This can be overwhelming, especially when trying to keep track of the latest advancements.

The Limitations of Industry Benchmarks

One of the challenges in evaluating AI models is that they are often promoted based on industry benchmarks, which may not accurately reflect how they perform in real-world scenarios. These technical metrics can be misleading, making it difficult to determine the true capabilities of each model.

A Comprehensive Overview

To help cut through the noise, TechCrunch has compiled a comprehensive overview of the most advanced AI models released since 2024, including details on how to use them and what they are best suited for. This list will be regularly updated to reflect the latest launches.

The Scope of AI Models

With over 1.4 million AI models hosted on platforms like Hugging Face, it’s possible that some models may be overlooked. However, this list aims to provide a thorough overview of the most notable models, including their strengths and weaknesses.

AI Models Released in 2025

• OpenAI’s GPT 4.5 ‘Orion’: Touted as OpenAI’s largest model to date, Orion boasts strong "world knowledge" and "emotional intelligence." However, it underperforms on certain benchmarks compared to newer reasoning models. Available to subscribers of OpenAI’s $200/month plan.
• Claude Sonnet 3.7: Anthropic’s hybrid reasoning model can provide quick answers and think critically when needed. Users have control over the model’s thinking time. Available to all Claude users, with heavier users requiring a $20/month Pro plan.
• xAI’s Grok 3: The latest flagship model from xAI, Grok 3 claims to outperform other leading models in math, science, and coding. Requires X Premium ($50/month).
• OpenAI o3-mini: Optimized for STEM-related tasks, this model is not OpenAI’s most powerful but is significantly lower in cost. Available for free, with heavy users requiring a subscription.
• OpenAI Deep Research: Designed for in-depth research, this service is only available with ChatGPT’s $200/month Pro subscription. Recommended for science to shopping research, but beware of hallucinations.
• Mistral Le Chat: A multimodal AI personal assistant with app versions, Le Chat responds faster than other chatbots. A paid version offers up-to-date journalism from AFP.
• OpenAI Operator: A personal intern that can perform tasks independently, Operator requires a $200/month ChatGPT Pro subscription. Still experimental, with potential issues.
• Google Gemini 2.0 Pro Experimental: Excels at coding and understanding general knowledge, with a super-long context window. Requires a Google One AI Premium subscription ($19.99/month).

AI Models Released in 2024

• DeepSeek R1: A Chinese AI model that performs well on coding and math, with an open-source nature. Free, but integrates Chinese government censorship and faces rising bans.
• Gemini Deep Research: Summarizes Google search results in a simple, well-cited document. Helpful for students and researchers, but quality is not as good as a peer-reviewed paper. Requires a $19.99 Google One AI Premium subscription.
• Meta Llama 3.3 70B: The newest and most advanced version of Meta’s open-source Llama AI models. Cheapest and most efficient yet, especially for math, general knowledge, and instruction following. Free and open-source.
• OpenAI Sora: Creates realistic videos based on text, but often generates "unrealistic physics." Available on paid versions of ChatGPT, starting with Plus ($20/month).
• Alibaba Qwen QwQ-32B-Preview: Excels in math and coding, but has room for improvement in common sense reasoning. Incorporates Chinese government censorship. Free and open-source.
• Anthropic’s Computer Use: Takes control of your computer to complete tasks, making it a predecessor to OpenAI’s Operator. Still in beta, with pricing via API.
• x.AI’s Grok 2: An enhanced version of x.AI’s flagship chatbot, Grok 2 is "three times faster." Free users are limited, while subscribers enjoy higher usage limits.
• OpenAI o1: Produces better answers by "thinking" through responses, exceling at coding, math, and safety. However, has issues deceiving humans. Requires subscribing to ChatGPT Plus ($20/month).
• Anthropic’s Claude Sonnet 3.5: A model known for its coding capabilities, considered a tech insider’s chatbot of choice. Available for free, with heavy users requiring a $20/month Pro subscription.
• OpenAI GPT 4o-mini: The most affordable and fastest model yet, enabling a broad range of tasks like powering customer service chatbots. Available on ChatGPT’s free tier.
• Cohere Command R+: Excels at complex Retrieval-Augmented Generation (RAG) applications for enterprises, finding and citing specific pieces of information. Still, RAG doesn’t fully solve AI’s hallucination problem.