MODEL

GPT-6 Luna

modeltopic-noteopenai

Overview

GPT-6 Luna is OpenAI‘s Sept 22, 2026 high-throughput, low-cost variant in the GPT-6 generation. Luna prices at $0.10/$0.50 per Mtok input/output with a 90% cached-input discount tier — roughly half the GPT-5.6 Luna counterpart. Distribution reaches free ChatGPT users and desktop, in contrast to GPT-6 Sol which lands in ChatGPT Work, Codex, and the API for paid tiers. Ships alongside GPT-6 Sol the same day as Anthropic‘s Claude Opus 5.5 release — the sub-24h paired frontier launch with matched per-token cuts is the coordinating datapoint.

Timeline

  • 2026-09-23-AI-Digest — OpenAI ships GPT-6 Luna as the high-throughput tier of the GPT-6 generation on Sept 22 — priced at $0.10/$0.50 per Mtok input/output with a 90% cached-input discount tier, roughly half GPT-5.6 Luna‘s counterpart. Luna reaches free ChatGPT users and desktop; sibling GPT-6 Sol lands in ChatGPT Work, Codex, and the API for paid tiers. Ships inside the 12-hour window with Anthropic‘s Claude Opus 5.5 release — Simon Willison flagged the sub-24h dual launch with matched per-token cuts as unprecedented.
  • 2026-09-24-AI-Digest — GPT-6 Luna’s $0.10/$0.50 per Mtok pricing is the anchor for the “Tokens too cheap to meter” HN post (269 pts · 188 cmts) — 2026 commodity-tier inference has fallen below the threshold where per-call accounting is worth its overhead, with implications for app and agent architecture. Load-bearing framing to carry: the claim is tier-specific, not market-wide — Luna at $0.10/$0.50 and Gemini 3.8 Flash TTS ($9/Mtok intro) live in the sub-cent commodity tier where per-call accounting is genuinely getting expensive to itself; Claude Opus 5.5 at $4/$20 and gpt-5 (high) reasoning tokens do not. Read the framing as a design signal for commodity-tier agents specifically.
  • 2026-10-08-AI-Digest — GPT-6 Luna is named as the only supported model for the new Decisions API in public beta, a structured-judgment endpoint OpenAI shipped alongside the GPT-6 drop. Per The Decoder (single secondary source), the Decisions API prices at $0.10/M input with output tokens free, delivers ~10× the speed of the Responses API with a three-type response shape (yes/no probabilities, category pick, scale rating), and the paid-tier structure collapses from five to three (Build at $500/mo, Launch at $5,000/mo, Grow at $200,000/mo). Load-bearing softener: the underlying developers.openai.com docs were not independently reachable — treat Decisions API pricing and tier structure as single-sourced. The $0.10/M input price now matches Anthropic‘s Claude Haiku 5.5 ≤100K input tier shipped hours earlier — small-tier input-price floor is now effectively zero-differentiated across the two frontier US labs. Log against MOC - Major Companies and MOC - Developer Tools.

Key Developments

  1. Priced at $0.10/$0.50 With 90% Cached-Input Discount (September 22, 2026): Standard pricing lands at roughly half the GPT-5.6 Luna counterpart, with the cached-input discount tier at 90% doing much of the load-bearing work for prompt-cache-heavy agent workloads. The tier structure mirrors Claude Opus 5.5‘s cache-heavy pricing shape — both labs cutting the same day on cache-read economics is the substrate story.

  2. Free-Tier Distribution as the Structural Move (September 22, 2026): Luna reaches free ChatGPT users and desktop while GPT-6 Sol carries the reasoning premium on paid tiers — continues the top-of-funnel-broadening pattern the GPT-5.6 Luna free-tier move started (July 31 80% cut, August 6 Luna-to-free-tier). Free-tier compression of the free-vs-paid boundary now extends to the GPT-6 generation.

  3. Same-Day With Sol as Paired Frontier Launch Against Claude Opus 5.5 (September 22, 2026): The sub-24h paired-launch framing carries to Luna: GPT-6 Sol at $2/$10, GPT-6 Luna at $0.10/$0.50, Claude Opus 5.5 at $4/$20 — all landing within 24 hours with matched per-token cuts. Load-bearing corpus reframe: coordinating datapoint on the price curve, not price war ignition.

See also: OpenAI, GPT-6, GPT-5.6 Luna, GPT-6 Sol, Astra, Claude Opus 5.5, Simon Willison, MOC - Major Companies, MOC - Agentic Coding.

  • 2026-10-22-AI-Digest — GPT-6 Luna entered private preview on Google Cloud on Oct 21 alongside GPT-6 Astra and GPT-6.1 Sol via Antigravity 2.0 and Gemini for Business. Enterprise buyers can apply existing Google Cloud spend commitments to the resulting OpenAI consumption on one GCP bill. OpenAI reiterated that Microsoft remains its primary cloud partner; GA is “in the weeks ahead” with no exact date; Codex coding tools were not confirmed in-scope by first-party coverage. Load-bearing framing: compute- and surfacing-channel diversification, not a strategic completion. Log against MOC - Major Companies.

  • 2026-11-03-AI-Digest — Simon Willison‘s llm 0.37 (2026-11-02) bumps the CLI’s default model to GPT-6 Luna. For teams running the llm toolchain, cron jobs and chained scripts that previously defaulted to the older model will silently route to Luna until pinned — a quiet but behaviour-changing switch on the 0.37 upgrade path. Pattern-matches the 0.32rc GPT-5.6 Luna default swap from 2026-07-31-AI-Digest, extending Willison’s practitioner-reroute-on-the-day-of-release pattern into the GPT-6 generation. Log against MOC - Developer Tools.