Skip to content
DopeSwagYolo

// artificial intelligence

AI Models: Releases, Comparisons and Explainers

Clear coverage of AI model releases, capabilities and pricing from OpenAI, Google, Anthropic, xAI and Meta, plus the open-weight models anyone can download and run.

Latest in AI Models

What we cover in AI Models

AI models are the engines behind chatbots such as ChatGPT, Gemini, Claude and Grok. They are also behind a growing set of open-weight systems that anyone can download. This section tracks new model releases, what each model can do, what it costs, and how the major labs compare. The major labs are OpenAI, Google, Anthropic, xAI and Meta. Open-weight developers such as DeepSeek, Qwen and Mistral are compared alongside them.

The pace of change is the reason this beat exists. All of this happened in September 2026 alone:

  • Anthropic released Claude Fable 5.1 and Claude Opus 5.5.
  • OpenAI released GPT-6 Astra and three lower-cost GPT-6 models.
  • Google shipped Gemini 3.8 Flash and announced Gemini 4 Argon.
  • xAI released Grok 4.7.

Rankings move as well. Stanford's 2026 AI Index reports that, as of March 2026, four companies sat within 25 points of one another on the Arena leaderboard. That leaderboard ranks models by human votes. The Index also reports that competitive pressure was moving toward cost and reliability. Arena's update of October 2, 2026 showed a wider spread. It showed a preliminary score for Gemini 4 Argon about 20 points ahead of the next entry.

Usage is shifting too. Estimates from Similarweb, a web analytics company, cover worldwide visits to generative AI websites. They show ChatGPT's share of those visits falling from about 76% in June 2025 to roughly 53% by May 2026. They show Gemini and Claude gaining ground over the same period. Because lineups change monthly, articles here state which model versions were current on the date they were checked. They attribute every benchmark result to whoever published it. No single model is treated as the best at everything. Agent products, regulation, chips and jobs are covered in other sections.

AI Models reference

AI Models: common questions

Which AI model is the best right now?
No single model leads everywhere. On the Arena text leaderboard updated October 2, 2026, Google's Gemini 4 Argon ranked first with a preliminary score. Several Anthropic Claude entries followed it. But Google says Argon is going to a limited group of users first. Benchmark tables published by the labs show different leaders for different tasks. The best choice depends on the job, the budget and the apps a person already uses.
Which AI model is best for coding?
It depends on the test. Google DeepMind published a comparison table for Gemini 4 Argon. In it, Argon led the DeepSWE v1.1 coding benchmark at 77.9%. In the same table, OpenAI's GPT-6 Astra led FrontierSWE v2 at 65.5% and Anthropic's Claude Opus 5.5 led Terminal-Bench 4.0 at 66.4%. These are vendor-published figures, so they are best read as claims rather than a neutral ranking.
Is DeepSeek open source?
DeepSeek publishes model weights under an open license. That is not the same as meeting the full definition of open source. The Hugging Face page for DeepSeek-V4.1-Flash says the repository and model weights are licensed under the MIT License. The Open Source Initiative's definition of open source AI asks for more than weights. That includes training code and detailed information about training data. A release of weights alone is what the group calls open weights.
Can I run AI locally on my PC?
Yes, if the model fits in the computer's memory. OpenAI says its open-weight gpt-oss-20b model runs within 16GB of memory. Google's documentation lists approximate memory needs for loading its Gemma 4 models. The listed needs range from 2.9GB to 17.5GB when the models are compressed to 4-bit precision, with long prompts needing more. Model pages on Hugging Face point to tools such as Ollama, LM Studio and llama.cpp for running models on personal hardware.

Sources

  1. Technical Performance | The 2026 AI Index Report | Stanford HAI, Stanford HAI
  2. AI Search Stats 2026: Market Share, Referral, and Citation Data, Similarweb
  3. Claude Fable 5.1 - Claude Platform Docs, Anthropic
  4. Introducing Claude Opus 5.5, Anthropic
  5. Changelog | OpenAI API, OpenAI
  6. Introducing Gemini 3.8 Flash and 3.8 Flash Cyber, Google
  7. Gemini 4 Argon: our next era of frontier intelligence, Google
  8. Introducing Grok 4.7, SpaceXAI (xAI)
  9. Text Arena: LLM Rankings & AI Model Benchmarks, Arena
  10. Gemini - Google DeepMind, Google DeepMind
  11. deepseek-ai/DeepSeek-V4.1-Flash · Hugging Face, DeepSeek, via Hugging Face
  12. The Open Source AI Definition – 1.0, Open Source Initiative
  13. Open Weights: not quite what you've been told, Open Source Initiative
  14. openai/gpt-oss-20b · Hugging Face, OpenAI, via Hugging Face
  15. Gemma 4 model overview | Google AI for Developers, Google
  16. Qwen/Qwen3.8-27B · Hugging Face, Qwen, via Hugging Face