Enterprise

OpenAI cuts GPT-6 Sol and Luna API prices in half, resetting the cost floor for AI sales and marketing tools

The permanent 50% API cut lands the same afternoon Anthropic ships Opus 5.5 — and lowers the operating cost of every agentic sales, marketing, and content tool small teams already pay for.

By Asher Glick· Enterprise software desk·2026-09-22·3 min read

OpenAI released GPT-6 Sol and GPT-6 Luna on Tuesday with API prices roughly 50% below the previous 5.6 generation, and an OpenAI spokesperson confirmed to VentureBeat that the new rates are permanent, not promotional. Sol lands at $2 per million input tokens and $10 per million output tokens; Luna at $0.10 and $0.50. OpenAI attributes the reduction to caching and inference improvements, and is nudging buyers toward a different unit of measurement: cost per completed task rather than tokens alone.

The framing matters. On AutomationBench 1.0.6, OpenAI’s new agent benchmark that tests end-to-end workflows across 47 tools spanning sales, marketing, operations, support, finance, and HR, Sol at extra-high effort outperforms Claude Opus 5 at maximum effort at 9% of its cost per task. Cached input reads now carry a 90% discount, and OpenAI says it has raised default cache hit rates.

The competitive choreography was tight. Anthropic shipped Claude Opus 5.5 roughly 90 minutes before OpenAI’s announcement, claiming Opus 5.5 costs about 40% less to run than Opus 5 on typical workloads. Yesterday, Jensen Huang pushed back on frontier-AI pacing concerns in a way that pointed to exactly this dynamic: the frontier isn’t slowing, it’s compressing on price.

For a five-to-thirty-person business, the practical read isn’t about which lab won the afternoon. It’s that the AI-native vendor stack sitting inside customer-acquisition work, outbound tools, CRM copilots, content generators, the ad platforms behind ChatGPT’s new business discovery surface, just saw its raw operating cost drop. Whether that shows up as lower subscription prices, higher usage caps, or better default output is a vendor-by-vendor question owners should be asking this week.

Routing logic is also shifting. Luna’s pricing fits bulk work: extraction, classification, summarization at volume. Sol targets higher-judgment coding and agent tasks at Sonnet-class token economics. Both are available through the ChatGPT API, ChatGPT Work, and Codex, and GitHub is rolling Sol into Copilot’s Pro+, Max, Business, and Enterprise plans, with Luna reaching Pro as well. Luna is also arriving in the ChatGPT desktop app for Free and Go users.

Every prior price war at the model layer ended with the savings captured somewhere between the lab, the vendor, and the customer. Small-team owners renewing contracts this quarter are the ones who decide where.

Sources