← All guides

GPT-5.6 Pricing: Sol, Terra & Luna API Cost Compared

Published 2026-06-27 · 6 min read

GPT-5.6 is OpenAI's newest model family, announced on 26 June 2026 — two months after GPT-5.5. Instead of one flagship it ships as three named tiers: Sol, Terra and Luna. This is the GPT-5.6 pricing breakdown — what each tier costs per token, a real worked bill, and how it compares to GPT-5.5, Claude and Gemini.

Availability: GPT-5.6 launched in limited preview (a small set of partner organisations), with general availability expected in the coming weeks. The numbers below are the announced list prices per 1M tokens — always confirm on OpenAI's official pricing page before committing to volume.

GPT-5.6 API pricing per 1M tokens

ModelInput /1MOutput /1MCached inputBest for
GPT-5.6 Sol$5.00$30.00$0.50Flagship — hardest coding & reasoning
GPT-5.6 Terra$2.50$15.00$0.25Daily driver — high-volume work
GPT-5.6 Luna$1.00$6.00$0.10Budget frontier — drafting, summarising

All three share a 400K-token context window and bill cached input at 90% off (the GPT-5.x family standard). Output costs 6× the input rate, so the length of the model's replies drives most of your GPT-5.6 bill.

What a real GPT-5.6 bill looks like

Take a support assistant: an 1,800-token prompt (instructions + context) and a 400-token answer, run 200,000 times a month.

  • Sol — (1,800×$5 + 400×$30) / 1M × 200K = ~$4,200/mo
  • Terra~$2,100/mo (exactly half of Sol)
  • Luna~$840/mo

Same workload, a 5× spread across the lineup. Plug your own volumes into the cost calculator to see where you land.

GPT-5.6 vs GPT-5.5, Claude and Gemini

For that same 1,800-in / 400-out workload at 200K calls per month:

ModelInput /1MOutput /1MThis workload
GPT-5.6 Sol$5.00$30.00~$4,200/mo
GPT-5.5$5.00$30.00~$4,200/mo
GPT-5.6 Terra$2.50$15.00~$2,100/mo
Claude Fable 5$10.00$50.00~$7,600/mo
Claude Opus 4.8$5.00$25.00~$3,800/mo
Gemini 2.5 Pro$1.25$10.00~$1,250/mo
GPT-5.6 Luna$1.00$6.00~$840/mo

Two takeaways. Sol costs exactly what GPT-5.5 did ($5/$30) but is the stronger model, so upgrading is a free win on capability. And Luna undercuts every flagship-class rival here — even Gemini 2.5 Pro — making it the cheapest frontier option for everyday work. For the upgrade decision in full, see GPT-5.6 vs GPT-5.5.

Cut the bill with prompt caching

If you resend the same system prompt or document on every call, GPT-5.6 bills those cached input tokens at 90% off — $0.50/1M on Sol, $0.25 on Terra, $0.10 on Luna. For RAG and agent workloads where context repeats, caching is the single biggest lever on a GPT-5.6 bill. (See why output tokens cost more for the other half of the equation.)

The bottom line on GPT-5.6 pricing

Pick Sol for the hardest agentic and coding work, Terra as the default daily driver at half the flagship price, and Luna when you want frontier quality at the lowest cost. Compare all three against every other model on the live pricing table, or open each tier's page: Sol pricing, Terra pricing, Luna pricing.

Compare every model's price in the live pricing table, or estimate your own bill with the cost calculator.