OpenAI Codex Boosts Cache Reuse Across Agents

This is an infrastructure upgrade with direct developer impact: better caching can make parallel agent workflows faster and substantially more economical, even without changing the underlying model.

• –Shared context across subagents reduces redundant prompt processing in multi-agent tasks
• –Longer cache retention should make pauses, tool calls, and background work less punitive
• –Improved routing targets cache misses caused by requests landing on incompatible infrastructure
• –The benefits are greatest for long-running coding, research, and review workflows with repeated context
• –OpenAI has not published concrete before-and-after usage or cache-hit figures, so the real impact remains user-dependent