OpenAI: GPT-6 Luna Pro (batch)

GPT-6 Luna Pro (batch) is the batch-priced variant of OpenAI's GPT-6 Luna served with reasoning.mode set to pro. It uses the same underlying weights as GPT-6 Luna but allocates additional reasoning compute for harder multi-step work in coding, computer use, and agentic workflows, while batch processing cuts standard API list prices by half ($0.05 per million input tokens and $0.25 per million output tokens on typical routes). The model supports a 1.05 million-token context window, up to 128K output tokens, text and image inputs, and OpenAI Responses API tooling including function calling, web search, file search, computer use, and prompt caching.

Luna sits at the efficient end of the GPT-6 family introduced in September 2026 alongside GPT-6 Sol, inheriting alignment and capability improvements from GPT-6 Astra at substantially lower cost per token. Pro mode is aimed at production agents and Codex-style tasks where solution quality matters more than the lowest latency, without moving to Sol- or Astra-tier pricing. Batch delivery trades synchronous latency for economical throughput on large offline job queues.

Reported evaluations for the Luna family at high reasoning settings include strong multimodal and graduate-level science scores, competitive software-engineering agent results on DeepSWE and SWE-bench Verified, and solid long-context and terminal-agent numbers, while business workflow suites such as AutomationBench and Agents' Last Exam remain more challenging at the Luna tier. Because Luna Pro (batch) is an API serving configuration rather than a separate weight release, benchmark figures align with publicly reported GPT-6 Luna runs at comparable reasoning settings unless a provider publishes Pro-specific tables.