Sunday, Aug 2 | --:--
Back to home

OpenAI Cuts GPT-5.6 Luna Prices 80% and Terra 20%

On July 30, 2026, OpenAI cut API prices for GPT-5.6 Luna by 80% (to $0.20/$1.20 per million tokens) and Terra by 20% (to $2/$12), held Sol pricing steady, and added a Fast mode for Sol up to 2.5× speed at 2× standard price—crediting efficiency gains from Sol on its own serving stack.

Tech Insights Reporter 4 min read San Francisco, CA
Cover illustration for OpenAI Cuts GPT-5.6 Luna Prices 80% and Terra 20%

TLDR

OpenAI on July 30, 2026—three weeks after GPT-5.6 public launch—cut Luna API prices by 80% and Terra by 20%, left Sol list prices unchanged, and launched Fast mode for Sol in the API (up to 2.5× Standard speed at Standard price). Lower Luna/Terra rates also reduce how usage is counted against Codex and ChatGPT Work quotas. OpenAI credits Sol-assisted efficiency work: ~20% lower serving costs and 15%+ token-generation efficiency via speculative decoding improvements.

New API pricing (per 1M tokens)

Model Prior (approx. launch) July 30 Change
GPT-5.6 Sol $5 in / $30 out Unchanged
GPT-5.6 Terra $2.50 / $15 $2 / $12 −20%
GPT-5.6 Luna $1 / $6 $0.20 / $1.20 −80%
Sol Fast mode Standard price, up to 2.5× speed New SKU option

Primary: OpenAI index post “Advancing the price-performance frontier with GPT-5.6” and same-day X thread; numbers corroborated by CNBC and VentureBeat.

Other product notes (same announcement)

  • Auto-review in ChatGPT app and Codex CLI upgraded from GPT-5.4 → GPT-5.6 Luna; with new Luna price, OpenAI expects Auto-review ~10× cheaper.
  • Subscription prices/quota budgets for ChatGPT/Codex unchanged; Terra/Luna consume fewer credits per unit of work.
  • AWS pricing rollout noted as beginning the same day.

Competitive pressure

OpenAI frames the move as mission-aligned abundance of intelligence—but the market context is a price war: Chinese open-weight volume (Kimi K3, DeepSeek tiers), Anthropic Opus 5 cost positioning vs Fable, Google Gemini 3.6 Flash efficiency claims, and enterprise pushback on unconstrained token spend (CNBC). Cutting Luna hardest defends the high-volume tier while holding Sol as premium.

Why this story matters

Pricing is now a first-class model story, not a footnote. Repricing two of three GPT-5.6 tiers within three weeks of launch shows how fast cost curves move when open weights and rival Flash/Sonnet lines undercut US list prices. Developers should re-benchmark Luna/Terra vs Sol for agentic workloads; finance teams should re-model API and Work/Codex burn rates. Watch whether Sol eventually drops, how Fast mode affects quality/latency tradeoffs, and whether Anthropic/Google match within days.

Sources

Prior Coverage

Earlier Times of AI reporting on this thread.

Scroll to continue reading