OpenAI and Anthropic Release Rival Coding Models in the Same Week, Then Buy Competing Super Bowl Ads

Claude Opus 4.6 and GPT-5.3-Codex landed days apart, both aimed at software work, as the two companies' executives traded public insults and took their fight to the biggest advertising stage in America.

Anthropic released Claude Opus 4.6 and OpenAI released GPT-5.3-Codex within days of each other this week, both aimed at the same thing: writing and maintaining software. Opus 4.6 plans more carefully, keeps working at long tasks without losing the thread, handles much larger amounts of code at once, and is better at catching its own mistakes. GPT-5.3-Codex searches through codebases, runs commands, and debugs. The two are close enough in purpose that the comparisons wrote themselves.

Then, tonight, both companies ran advertisements during the Super Bowl — the most expensive airtime in American media, watched by more than a hundred million people. Their executives have spent the week trading barbs in public over business models, who gets access to what, and each other's corporate ethics.

There is something worth noting in the sequence. Three years ago these were research organizations that published papers and worried aloud about releasing things. This week they are two consumer brands buying Super Bowl spots to insult each other. That is a normal arc for a technology industry, and a strange one for an industry whose leaders signed a statement in 2023 saying their work might pose a risk of human extinction.

The reason both are spending on software specifically is that coding is where AI currently earns its keep. It is the task these systems do best, the one with the clearest measure of success, and the one customers pay for most readily. The Super Bowl ads are aimed at the much larger group of people who have not yet worked out what any of this is for.