Anthropic's 48-hour product blitz proves it's no longer an AI startup with a chatbot—it's becoming the AI platform the enterprise never had. Between Thursday and Saturday last week, the company shipped Claude Opus 4.7, Claude Design, and Claude Mythos Preview. Three launches. Three different markets. One unmistakable message.
Opus 4.7 leads the charge. The flagship model delivers strictly better performance than Opus 4.6 across all effort tiers—a rare achievement in incremental model development. The 4.7-low tier outperforms 4.6-medium, 4.7-medium eclipses 4.6-high, and 4.7-high surpasses 4.6-max. But raw capability gains only tell part of the story. Claude Code now defaults to xhigh effort level, pushing SWE-Bench Pro scores 11 points higher. More impressively, overall token consumption dropped by up to 50% despite a 35% increase from the new tokenizer—the reasoning efficiency gains are substantial. Vision capabilities now accept images up to 3.75 megapixels, roughly three times what prior Claude models could handle, enabling pixel-perfect analysis of dense screenshots and complex diagrams. And pricing remains anchored at $5 per million input tokens and $25 per million output tokens, exactly where Opus 4.6 launched.
This is the technical foundation. But the real story isn't Opus 4.7 alone—it's what Anthropic built around it.
Claude Design targets a new buyer entirely: founders, product managers, and others without design backgrounds who need to communicate visual ideas quickly. The tool enables rapid visual prototyping without requiring artistic skill. It's a direct product, not an API, marking Anthropic's first serious move into subscription revenue beyond developer tooling.
Claude Mythos Preview enters the cybersecurity and government space—a market Anthropic burned bridges with during its public standoff with the Trump administration over domestic surveillance and lethal autonomous weapons. The Mythos model appears designed to rebuild those bridges while keeping the company's stated ethical positions intact.
The 48-hour cadence is deliberate. Three products. Three audiences. One signal: Anthropic can operate a multi-product business at scale. The company that defined itself as a "model company" is now showing it can be a platform.
For developers, Opus 4.7 delivers more power at the same cost. For non-technical users, Design offers a direct path to AI-assisted visual work. For government customers, Mythos provides a potential on-ramp. The three-launch window was Anthropic's proof of concept for platform thinking. The next test is execution.