AI models are being produced at a rapid pace by a range of companies, from major technology firms such as Google to startups like OpenAI and Anthropic. This proliferation can make it challenging to keep track of the latest developments.
Contributing to the complexity is the fact that AI models are often marketed based on industry benchmarks, which may not accurately reflect their practical applications or usability for companies and individuals.
To streamline understanding, TechCrunch has curated a summary of the most advanced AI models launched since 2024, including how they can be utilized and their optimal purposes. This list will be regularly updated as new models are released.
There are over a million AI models available, with platforms like Hugging Face hosting more than 1.4 million. As a result, some models with superior performance in specific areas may not be included in this list.
AI Models Released in 2025
Google Gemini 2.5
Google’s Gemini 2.5 Pro Experimental is a reasoning model that excels in building web applications and code agents. However, it falls short on a certain coding benchmark compared to Claude Sonnet 3.7. Users can access this model with a $20 monthly Gemini Advanced subscription.
ChatGPT-4o Image Generator
OpenAI has enhanced its GPT-4o model to generate images in addition to text. This feature gained popularity for creating Studio Ghibli-style anime images, although copyright issues have been noted. Access to GPT-4o requires a minimum $20 per month ChatGPT Plus subscription.
Stability AI’s Stable Virtual Camera
Stability AI, a startup focusing on image generation, has released a model capable of generating 3D scenes and camera angles from a single 2D image. It struggles with complex elements like humans and moving water. Available for noncommercial research use via Hugging Face.
Cohere’s Aya Vision
Cohere introduced Aya Vision, a multimodal model that excels in captioning images and answering photo-related questions. It performs well in multiple languages, diverging from other models, and is free to use on WhatsApp.
OpenAI’s GPT 4.5 “Orion”
Touted by OpenAI as their largest model, Orion offers strong "world knowledge" and "emotional intelligence," though it underperforms on newer reasoning benchmarks. Orion is available for $200-per-month OpenAI subscribers.
Claude Sonnet 3.7
Anthropic’s Claude Sonnet 3.7 is described as the first "hybrid" reasoning model, which provides rapid answers while allowing thorough deliberation. Users can regulate the model’s response duration. It’s accessible to all Claude users, with a $20 monthly Pro plan for heavy usage.
xAI’s Grok 3
The latest flagship model from Elon Musk-founded xAI, Grok 3 is reported to surpass leading models in math, science, and coding. It requires an X Premium subscription ($50 per month). Musk has pledged to make Grok politically neutral after Grok 2 showed a left-leaning bias.
OpenAI o3-mini
OpenAI’s o3-mini is adapted for STEM tasks like coding, math, and science, providing a cost-effective, albeit less powerful, solution. It’s free but requires a subscription for extensive use.
OpenAI Deep Research
Targeted for in-depth research with clear citations, OpenAI’s Deep Research is part of ChatGPT’s $200-per-month Pro subscription. While useful for various types of research, the model is cautioned for potential hallucinations.
Mistral Le Chat
Mistral has developed Le Chat, a multimodal personal AI assistant. It reportedly has rapid response times and a paid version with current AFP journalism. Although testing praises its performance, it made more errors than ChatGPT.
OpenAI Operator
Designed as a personal assistant, OpenAI’s Operator independently performs tasks like grocery shopping. This service requires a $200 monthly ChatGPT Pro subscription. Despite its potential, it’s still in experimental stages, as noted by a Washington Post reviewer.
Google Gemini 2.0 Pro Experimental
The anticipated Google Gemini 2.0 Pro excels in coding and general knowledge understanding. Featuring a context window of 2 million tokens, it aids users in processing extensive text. It’s available with a Google One AI Premium subscription of $19.99 per month.
AI Models Released in 2024
DeepSeek R1
This Chinese AI model gained popularity in Silicon Valley due to its performance in coding and math. Its open-source nature allows local execution, although it incorporates Chinese government censorship and raises concerns about data privacy.
Gemini Deep Research
Providing summaries of Google search results, Gemini Deep Research simplifies information with proper citations. While beneficial for quick research, it doesn’t match peer-reviewed papers in quality. It requires a $19.99 Google One AI Premium subscription.
Meta Llama 3.3 70B
Meta’s latest Llama AI model is both cost-effective and efficient, particularly in math and general knowledge. As an open-source tool, it is freely available.
OpenAI Sora
Sora generates realistic videos from text, producing entire scenes. Despite its capabilities, OpenAI acknowledges it may deliver unrealistic physics. Available with ChatGPT’s paid tiers, starting at $20 monthly.
Alibaba Qwen QwQ-32B-Preview
Rivaling OpenAI’s o1 in benchmarks, it excels in math and coding yet needs improvement in common sense reasoning. It integrates Chinese government censorship, is free, and open source.
Anthropic’s Computer Use
An early version of a computer controlling model, Claude’s Computer Use manages tasks like coding. Still in beta and differing pricing via API, it’s a precursor to OpenAI’s Operator.
xAI’s Grok 2
Enhancing the previous version, Grok 2 is three times faster, with free users limited in capacity, whilst those on Premium plans have broader access. It also incorporates Aurora, an image generator with photorealistic outputs.
OpenAI o1
Designed for thorough responses, OpenAI’s o1 excels in coding, math, and safety but holds controversial tendencies in tricking users. It’s part of the ChatGPT Plus, priced at $20 monthly.
Anthropic’s Claude Sonnet 3.5
Regarded as premier in its class by Anthropic, Claude Sonnet 3.5 is favored for its coding capabilities among tech insiders. Free access is provided, though Pro subscriptions are offered at $20 per month for intensive use.
OpenAI GPT 4o-mini
Characterized by affordability and speed, GPT 4o-mini facilitates a range of simple, high-volume tasks, although more complex requirements are beyond its scope. It is accessible on ChatGPT’s free tier.
Cohere Command R+
Specialized in retrieval-augmented generation (RAG) applications for enterprises, Cohere’s Command R+ excels in finding and citing specific data effectively, though it doesn’t completely eliminate AI hallucinations.
