svper.ai #3
🏁 Anthropic & OpenAI Ship Flagship Models on the Same Day
Anthropic released Claude Opus 4.6 on February 5 — its most powerful model yet. The standout feature is "agent teams": multiple agents that split larger tasks, each owning its piece and coordinating directly with the others. During testing, Opus 4.6 found over 500 previously undisclosed zero-day vulnerabilities in well-tested open-source libraries. A fast mode preview followed on February 7, delivering up to 2.5x faster output.
TechCrunchOpenAI launched GPT-5.3-Codex the same day, combining the Codex and GPT-5 training stacks into one unified model that's 25% faster. It sets new highs on SWE-Bench Pro and Terminal-Bench, and shows strong performance on OSWorld and GDPval — four benchmarks used to measure coding, agentic, and real-world capabilities.
OpenAIAnthropic's zero-day discoveries came from codebases that had fuzzers running against them for millions of CPU hours — some vulnerabilities had gone undetected for decades. On the other side, OpenAI delayed full developer API access for GPT-5.3-Codex because it's the first model to hit "high" on their preparedness framework for cybersecurity.
FortuneThe same-day launches signal both companies are moving beyond chat interfaces toward autonomous agent platforms. Anthropic's agent teams let you split work across frontend, API, and migration agents. OpenAI launched Frontier, an enterprise platform for building, deploying, and managing agents with shared context, onboarding, and permissions.
TechCrunch