# OpenAI and Anthropic New Model Releases (They Are At War) Source: https://www.youtube.com/watch?v=9f2egsZZjnw Recap page: https://rapidrecap.app/video/9f2egsZZjnw Generated: 2026-02-05T23:04:56.302+00:00 --- ## Quick Overview Anthropic's Claude Opus 4.6 is shown to outperform OpenAI's GPT-5.3 Codex in knowledge work (90.4% vs 86.0%), but GPT-5.3 Codex leads in agentic coding (77.3% vs 69.4%), illustrating a competitive dynamic where both AI models launch major updates near the Super Bowl, forcing consumers to decide between a free, ad-supported OpenAI experience and Anthropic's ad-free, paid offerings. **Key Points:** - Anthropic's Claude Opus 4.6 achieved a higher score than GPT-5.3 Codex on the GPQA knowledge work benchmark (90.4% vs 86.0%). - GPT-5.3 Codex outperformed Claude Opus 4.6 in agentic coding (77.3% vs 69.4%) and agentic search (79.9% vs 86.0% accuracy, though the video context implies Codex is better here based on the comparison table reading). - Anthropic released Claude Opus 4.6 just 15 minutes before OpenAI released GPT-5.3 Codex, suggesting a direct competitive response. - Anthropic's advertising strategy emphasizes being ad-free, contrasting with OpenAI's plan to run ads in ChatGPT, which Anthropic frames as a potential consumer negative. - The video compares the user experience of coding tasks in both the Cursor IDE (using GPT-5.3 Codex) and Claude's interface, noting Claude's output looked cleaner initially. - Claude 4.6 introduces features like improved reasoning, a 1M token context window, and advanced agentic capabilities like 'adaptive thinking' and 'new effort controls'. ![Screenshot at 00:14: The host displays the NDTV Profit article headline comparing the OpenAI vs Anthropic Super Bowl ad beef, setting the stage for the competitive analysis of the two new AI models.](https://ss.rapidrecap.app/screens/9f2egsZZjnw/00-00-14.jpg) **Context:** The video discusses the recent, closely timed competitive launches of new flagship models from two leading AI companies: Anthropic (with Claude Opus 4.6) and OpenAI (with GPT-5.3 Codex). This competition intensified around the time of the Super Bowl, prompting commentary on the models' relative performance across various benchmarks and the differing monetization strategies (Anthropic remaining ad-free versus OpenAI introducing ads to its free tier). The speaker compares the models' capabilities, particularly in coding and knowledge work, using benchmark charts from the respective release announcements. ## Detailed Analysis The video analyzes the recent competitive releases of Claude Opus 4.6 by Anthropic and GPT-5.3 Codex by OpenAI, noting that Anthropic released their model just 15 minutes before OpenAI, suggesting a direct race to market. The speaker reviews benchmark data showing Claude Opus 4.6 leading in Knowledge Work (90.4% vs 86.0% on GPQA), while GPT-5.3 Codex appears superior in Agentic Coding (77.3% vs 69.4% on the Terminal Bench 3.0 equivalent shown in the comparison table) and Agentic Search. The competition extends to advertising: Anthropic's ads stress being ad-free, contrasting with OpenAI's move to run ads in ChatGPT's free tier, a strategy Anthropic implies is dishonest. The speaker then conducts a side-by-side test using a web design prompt in both the Cursor IDE (set to GPT-5.3 Codex) and Claude's interface (set to Opus 4.6). Claude's initial output appeared cleaner, while GPT-5.3 Codex's output was slightly behind in the first attempt. The speaker highlights new features in Claude 4.6, such as adaptive thinking and a 1M token context window, suggesting that the models are getting better at complex, multi-step reasoning and coding tasks, forcing continuous, rapid improvement across the industry. ### AI Model Releases & Competition - Anthropic released Claude Opus 4.6, followed closely by OpenAI's GPT-5.3 Codex, leading to an AI 'war' narrative. ### Benchmark Performance Comparison - Claude Opus 4.6 leads in Knowledge Work (90.4% vs 86.0% on GPQA), while GPT-5.3 Codex leads in Agentic Coding (77.3% vs 69.4% on Terminal Bench 3.0 equivalent). ### Advertising & Monetization War - Anthropic's ads emphasize being ad-free, contrasting with OpenAI's decision to introduce ads into ChatGPT's free tier, framing it as a key differentiator. ### Claude 4.6 New Features - Introduction of a 1M token context window, adaptive thinking capabilities, and new effort controls for developers. ### Side-by-Side Coding Test - A web design prompt was given to both models; Claude's initial output appeared cleaner, while GPT-5.3 Codex was slightly slower in generating the full result. ![Screenshot at 00:04: The speaker displays an NDTV Profit article discussing the 'Super Bowl AI Ad Beef' between OpenAI and Anthropic.](https://ss.rapidrecap.app/screens/9f2egsZZjnw/00-00-04.jpg) ![Screenshot at 00:29: A bar chart from GPTrends showing Monthly Unique Visitors, highlighting ChatGPT's dominance \(415.7M\) over Claude \(15.5M\), Gemini \(122.3M\), Deepseek \(50.1M\), and Perplexity \(19.0M\).](https://ss.rapidrecap.app/screens/9f2egsZZjnw/00-00-29.jpg) ![Screenshot at 01:32: Anthropic's announcement page for Claude Opus 4.6, detailing improvements to coding skills and a 1M token context window.](https://ss.rapidrecap.app/screens/9f2egsZZjnw/00-01-32.jpg) ![Screenshot at 01:39: OpenAI's announcement page for GPT-5.3-Codex, emphasizing its capabilities as an agentic coding model.](https://ss.rapidrecap.app/screens/9f2egsZZjnw/00-01-39.jpg) ![Screenshot at 02:01: A bar chart from the Anthropic announcement showing Opus 4.6 leading in 'Knowledge work' benchmarks against other frontier models.](https://ss.rapidrecap.app/screens/9f2egsZZjnw/00-02-01.jpg)