The hottest AI models, what do they do and how to use them
AI models are discharged at dizziness rates, from all of major technology companies such as Google to start -ups such as Openai and Anthropic. Tracking the most recent can be captivating.
The addition of confusion is that AI models are often promoted on the basis of indicators in the industry. But those Technical indicators often reveal a little How real people and companies actually use them.
To cross the noise, TechCrunch has been a review of the most modern AI models launched since 2024, with details of how to use them and what they are for the best. We will also support this list with the latest start.
There are literally over a million AI models there: a hug, for example, hosts over 1.4 millionS So this list can miss some models that perform better, in one way or another.
AI models issued in 2025
Google Gemini 2.5
Gemini 2.5 Pro Experimental, a model of reasoning, is distinguished by Build web applications and code agents According to Google. However, it is smaller than a popular coding indicator than Claude Sonnet 3.7, however. The model requires an extended subscription for twins from $ 20.
Chatgpt-4o Image Generator
OPENAI has upgraded its existing GPT-4o model To generate images, not just text. The Supply Model Soon became viral for Ghibli -style anime -style conversionDespite the obvious copyright problems. Access to the GPT-4O requires at least $ 20 per month subscription to Chatgpt plus.
Stability of a stable virtual camera of AI
Starting images to start images AI launched a model This says that the company can generate 3D scenes and corners of a 2D image. However, he is still struggling with scenes involving more complex elements such as people and moving water. The model is accessible to the use of non -commercial research on Huggingface.
Aya’s Coher’s Vision
Cohere released a multimodal model Called the Aya Vision, which he claims to be the best in class in doing things like an inscription and answering questions about photos. It also features languages other than English, unlike other models, Cohere claims. It is available for Free on WhatsAppS
GPT 4.5 of Openai 4.5
Openai calls Orion them The largest model so farby saying his strong “world knowledge” and “emotional intelligence”. However, it is smaller than certain indicators than the larger models of reasoning. Orion is available to Openai’s $ 200 per month subscribers.
Claude Sonnet 3.7
Anthrop says it’s The first model for “hybrid” reflection in industryBecause I can shoot quick answers and really think things when needed. It also gives users to control how long can the model think, anthropyS Sonnet 3.7 is available to all CLAude users, but heavier users will need a $ 20 per month Pro plan.
Gook 3 on xia
Grok 3 is The last leading model From the Elon Musk launcher for launching XAI. It’s it is claimed that It outperforms other leading models in mathematics, science and encoding. The model requires X Premium (which is $ 50 per month.) After one study found Grok 2 bent left, musk promised To shift Grock more “politically neutral”, but it is not yet clear whether this has been achieved.
OPENAI O3-Mini
It’s on Openai The last model of reasoning and is optimized for STEM tasks such as encoding, mathematics and science. It’s Is not the most powerful of Openai model but as it is smaller, the company He says This is a significantly lower price. It is available for free but requires a subscription for heavy users.
Openai Deep Research
Openai’s deep study is Designed to carry out in -depth research On a topic with clear quotes. This service is only available with Chatgpt’s Subscription for $ 200 per monthS OPENAI Recommended it For everything from science to shopping, but be careful Hallucinations remain a problem For AI.
Mistral the Cat
Mistral has Started LE Chat Apps versionsMultimodal AI personal assistant. Mistral claims Le Chat responds faster than any other chatbot. In addition, there is a paid version with Up -to -date journalism by AFP. Tests from the world I found the presentation of Le Chat impressive, although made more mistakes than Chatgpt.
OPENAI operator
Openai’s operator is intended to be A personal trainee who can do things independently by helping you buy groceries. It requires a Chatgpt Pro subscription of $ 200 per month. Ai agents have many promises, but they are still experimental: Washington Post reviewer says an operator He decided to order a dozen eggs for $ 31, paid with the reviewer’s credit card.
Google Gemini 2.0 Pro Experimental
Google Gemini’s Flagship He says he is distinguished by the encoding and understanding of common knowledge. It also has a super -long context window of 2 million tokens, helping users who need to quickly process massive pieces of text. The service requires (at least) Google One Ai Premium subscription of $ 19.99 per month.
AI models issued in 2024
Deepseek R1
This Chinese AI model took Silicon Valley from a stormS Deepseek R1 performs well in encoding and math, while his open source nature means that anyone can manage it locally. Plus it’s free. However, R1 is integrated Chinese government censorship and Persons Raising to potentially send user data back to China.
Twin studies
Deep research Summarizes Google search results in a simple and well quoted document. Service is useful for students and for anyone else who needs a quick summary of research. However, its quality is not as good as actual review paper. Deep research requires a $ 19.99 dollars subscription to a Google One Ai Premium premium.
Goals of Calls 3.3 70B
It’s The most recent and most modern version Meta Llama AI’s open code. Meta is advertised this version As its most cheapest and most effective so far, especially for mathematics, general knowledge and the following instructions. It is a free and open code.
OPENAI SORA
Sora is a model that Creates realistic videos based on text. Although it can generate whole scenes, not just videos, Openai recognizes that it often generates “unrealistic physics”. It is currently only available in paid versions of Chatgpt, starting with Plus, which is $ 20 a month.
Alibaba Qwen QWQ-32B-Preview
This model is One of the few that rivals on Openai’s O1 for certain indicators in the industry, distinguished by mathematics and encoding. Ironically, a “model of reasoning”, he has a “place to improve reasoning of common sense”, “” Alibaba saysS It also includes the censorship of the Chinese government, TechCrunch test testsS It is a free and open code.
Using an anthropic computer
Claude’s computer use is intended for Take control of your computer To accomplish tasks such as encoding or booking a plane ticket, which makes it a precursor to the OPENAI operator. However, using a computer, however, stays in betaS The pricing is via API: $ 0.80 per million tokens for entry and $ 4 per million tokens from production.
Xia’s Gook 2
AI Company on Elon Musk, XAI, launched Improved version of its lead Grok 2 Chatbot it claims is “three times faster.” Free users are limited to 10 questions every two hours of Grok, while PREMIUM and Premium Subscribers+ Plans of X are used by higher restrictions on use. XAI also released an image generator, Aurora, this produces highly photorealistic imagesincluding some graphic or violent content.
OPENAI O1
Openai’s O1 family aims to give better answers by “thinking” through answers through hidden Reflection functionS The model is superior to coding, mathematics and safety, OPENAI claimsbut there is Problems and attempts to mislead people. Using O1 requires a subscription to Chatgpt Plus, which is $ 20 per month.
Anthropic’s Claude Sonnet 3.5
Claude Sonnet 3.5 is an anthrop model Claims as best in classS He became known for his coding capabilities and is considered a technological insider Choice chatbot. The model can be available free of charge to CLAUDE, although heavy users will need a $ 20 monthly Pro subscription. Although he can understand the images, he cannot generate them.
OPENAI GPT 4O-Mini
Openai has advertised GPT 4o-Mini As the most affordable and fastest model so far, thanks to its small size. This is intended To activate a wide range of tasks such as powering customer service chatbots. The model is accessible to the free Chatgpt layer. It is more suitable for simple high volume tasks than more complex.
Cohere Command R+
Stomach Command r+ model Overburns the applications of complex extract applications (or RAG) for businesses. This means that he can find and cite specific parts of the information really well. (The inventor of RAG actually works in Cohere.) Still, a rag It does not completely solve the problem of hallucination of AI.