Android Bench 2.0 ranks GPT-6 Astra and Claude Fable 5.1 top with coding agents

Android Bench 2.0 evaluates model and coding agent combinations, such as GPT with Codex and Claude with Claude Code. GPT-6 Astra plus Codex and Claude Fable 5.1 plus Claude Code took the top two spots, with A and O companies sweeping the top four positions.