The builder’s guide to GPT‑5.6

GPT-5.6 Luna matches GPT-5.5's BrowseComp score for $1.33 versus $33.27, and new API settings triple ARC-AGI-3 scores.

• GPT-5.6 Sol at "low" reasoning outperforms GPT-5.5 at "high" reasoning on Agents' Last Exam.
• New Responses API primitives: persisted reasoning, native compaction, multi-agent orchestration, and programmatic tool calling.
• Prompt cache TTL extended to a minimum of 30 minutes with deterministic cache breakpoints.
• Hypha reports Luna keeps 98% of GPT-5.5's extraction accuracy at one-eighteenth the cost.

Startups building agents can now get frontier-level performance at a fraction of the cost, changing the economics of agentic workflows.

Sources

• OpenAI: The builder’s guide to GPT‑5.6

Read this as text

Back to the AI news