GPT-6 Sol & Luna: OpenAI Cuts API Prices ~50% vs GPT-5.6 Promo

OpenAI ships cheaper GPT-6 workhorses

On September 22, 2026, OpenAI introduced GPT-6 Sol and GPT-6 Luna — faster, more affordable GPT-6 models trained with methods similar to GPT-6 Astra. The headline for builders: API prices for both are about 50% below GPT-5.6 promotional pricing, and the models are live in ChatGPT Work, Codex, and the API today.

This page is the long-form briefing behind the BSH Technologies Instagram carousel, with primary OpenAI sources linked throughout.

What happened

Sol and Luna expand the GPT-6 family after Astra. OpenAI frames them as everyday workhorse models that bring much of Astra's strength into cheaper, faster options for work at scale — including coding agents and long agent loops. Caching and inference efficiency improvements are part of the same release; OpenAI says it is passing those savings through as the price cut.

• Primary announcement: OpenAI — Introducing GPT-6 Sol and Luna
• API model card (Sol): developers.openai.com — gpt-6-sol
• API model card (Luna): developers.openai.com — gpt-6-luna

API pricing (per 1M tokens)

GPT-6 Sol — mid-tier workhorse

• Input: $2.00
• Output: $10.00
• API ID: gpt-6-sol
• Compared with GPT-5.6 Sol promo pricing ($4 / $20), that is a clean ~50% cut on both input and output.

GPT-6 Luna — high-volume efficient tier

• Input: $0.10
• Output: $0.50
• API ID: gpt-6-luna
• Compared with GPT-5.6 Luna promo pricing ($0.20 / $1.20), input is half and output is also sharply lower.

Prompt caching is priced to encourage agent loops: cached input is billed at about 10% of the uncached input rate on both models (Sol cached input ~$0.20; Luna cached input ~$0.01 per 1M). That matters if you re-send large system prompts or tool schemas on every turn.

Who can use it today

• ChatGPT Work + Codex: Sol and Luna roll out for Plus, Pro, Business, Enterprise, and Edu.
• API: both models available with IDs gpt-6-sol and gpt-6-luna.
• Free / Go: Luna available in the desktop app (gradual rollout).

When to pick Sol vs Luna vs Astra

• Luna — focused, high-volume tasks where $0.10 input keeps agent or batch cost low.
• Sol — complex coding and agentic workflows that need more capability than Luna without Astra's top-tier bill.
• Astra — still the frontier GPT-6 option when quality outweighs unit cost.

Why it matters

• Usage-limit math shifts. Same budget buys roughly 2× tokens on the Sol path vs GPT-5.6 Sol promo rates.
• Agent loops get cheaper. Lower list prices plus cheaper cached reads reduce the tax on long tool-using sessions.
• Product surface is broad on day one. Work, Codex, and API availability means teams can try the same IDs in chat and in production code without waiting for a second launch wave.

Primary sources

• OpenAI — Introducing GPT-6 Sol and Luna
• OpenAI API docs — GPT-6 Sol
• OpenAI API docs — GPT-6 Luna
• OpenAI API docs — Pricing

How BSH can help

At BSH Technologies we help teams pick model tiers for coding agents, evaluate cost/quality tradeoffs, and turn vendor launch notes into decision memos your stakeholders can trust — with sources. If you are deciding when to move workloads from GPT-5.6 promo SKUs to Sol or Luna (or keep Astra for the hard path), our Thrissur engineers can map that against your token mix and SLAs.