Claude Opus 4.6 adds agent teams, 1M context, and fast mode; GPT-5.3-Codex counters
Anthropic’s Claude Opus 4.6 ships multi-agent coding, a 1M-token context window, and a 2.5x fast mode, while OpenAI’s GPT-5.3-Codex brings faster agentic coding with strong benchmark results. DeepLearning.ai details Opus 4.6’s long-context, agentic coding gains, new API controls, and Codex 5.3’s speed and scores, plus pricing context [Data Points: Claude Opus 4.6 pushes the envelope](https://www.deeplearning.ai/the-batch/claude-opus-4-6-pushes-the-envelope/)[^1]. AI Collective highlights Claude Code’s new multi-agent “agent teams,” Office sidebars, and head-to-head benchmark moves versus OpenAI, while Storyboard18 confirms a 2.5x “fast mode” rollout for urgent work [Anthropic’s Opus 4.6 Agent Teams & OpenAI’s Codex 5.3](https://aicollective.substack.com/p/the-brief-anthropics-opus-46-agent)[^2] and [Anthropic rolls out fast mode for Claude Code](https://www.storyboard18.com/digital/anthropic-rolls-out-fast-mode-for-claude-code-to-speed-up-developer-workflows-89148.htm)[^3]. [^1]: Roundup covering features, benchmarks, and pricing for Opus 4.6 and GPT‑5.3‑Codex. [^2]: Newsletter with details on "agent teams," 1M-context performance, Office integrations, and comparative benchmarks. [^3]: Report on the 2.5x faster "fast mode" availability and target use cases.