GPT-6 Sol is now an official OpenAI model: id gpt-6-sol, released alongside GPT-6 Luna
From community label to official model
On 2026-09-22 OpenAI published "Introducing GPT-6 Sol and Luna". The earliest source this page can trace for the name is a Feature requests thread on the OpenAI developer community from 2026-09-11; it is now an official model: 1,050,000-token context, 128,000 max output, $2 / $10 input / output. On QCode, keep the same key and set model to gpt-6-sol.
Updated 2026-09-19
Four key points
Official release date
OpenAI's news feed published "Introducing GPT-6 Sol and Luna" — one announcement for both models.
Official model id
Listed on the official pricing page, next to gpt-6-luna and gpt-6-astra.
Input / output (per 1M tokens)
Cached input $0.20. Calculated from the official price table, that is half of gpt-5.6-sol ($4 / $20).
Context window (tokens)
128,000 max output tokens, knowledge cutoff 2026-04-20.
What this is
GPT-6 Sol and GPT-6 Luna are two models OpenAI released in one announcement on 2026-09-22; OpenAI describes them as different balances of capability and cost. Sol's official id is gpt-6-sol, Luna's is gpt-6-luna; the higher-priced gpt-6-astra ($10 / $50) remains. "Terra", listed next to them in the community thread, is not in the announcement's title.
What happened
On 2026-09-22 OpenAI's official news RSS carried "Introducing GPT-6 Sol and Luna" and then "Better prompt caching for GPT-6" — the latter covers higher cache hit rates, new diagnostics, explicit breakpoints, and controls that reduce latency and costs.
Timeline
2026-09-11 to 2026-09-12 the OpenAI developer community thread "GPT-6 Sol, Terra, and Luna?" discusses the names under Feature requests — the earliest source this page can trace.
The 2026-09-12 official pricing archive does not yet list gpt-6-sol; the callable ids then were gpt-5.6-sol and gpt-6-astra.
2026-09-22: OpenAI publishes "Introducing GPT-6 Sol and Luna"; gpt-6-sol and gpt-6-luna become official models listed on the pricing page.
Confirmed vs caution
Officially confirmed
What official sources confirm: on 2026-09-22 OpenAI's news feed announced GPT-6 Sol and Luna; the Sol model page states a 1,050,000 context window, 128,000 max output tokens, a 2026-04-20 knowledge cutoff, input $2.00, cached input $0.20 and output $10.00; the official pricing page lists gpt-6-luna at $0.10 / $0.50, gpt-6-astra at $10 / $50 and gpt-5.6-sol at $4 / $20.
⚠️ Caution
OpenAI's 2026-09-22 launch post carries self-reported benchmarks; this page transcribes none of them and states GPT-6 Luna's context as on its official model page (checked 2026-09-29): 1,050,000 context window, 922,000 maximum input tokens, 128,000 max output tokens. "Half" is our own figure, calculated from the official price table, not OpenAI's wording. Above 272K input tokens the whole request moves to long-context pricing, so long-document work costs more than the $2 / $10 base rate. Status update: OpenAI released gpt-6.1-sol on 2026-09-29 in the same tier, with cached input priced at 5% instead of 10%. This page still tracks gpt-6-sol; gpt-6.1-sol has been callable on QCode since 2026-09-30.
Same vs differs
Same
gpt-6-sol and gpt-6-luna were released on the same day in the same official announcement, both are on the official pricing page, and both have a separate long-context price tier.
Differs
The prices are far apart: Sol is $2 / $10 input / output, Luna $0.10 / $0.50. Sol has a 1,050,000 context window and 128,000 max output; Luna's official model page gives 1,050,000 context and 922,000 maximum input (checked 2026-09-29).
What to do
On the same QCode key, set model to gpt-6-sol. For a lower unit price, try gpt-6-luna (official price $0.10 / $0.50). Watch the long-context tier: requests with more than 272K input tokens are billed by OpenAI at 2x input and cache rates and 1.5x output for the full request.
On QCode
QCode's /models now lists gpt-6-sol (recommended), gpt-6-luna, gpt-6-astra, gpt-5.6-sol and gpt-5.6-terra; switch by changing model on the same key. gpt-5.6-luna has been removed from the list; gpt-5.5 can still be called but is no longer listed and is not recommended for new setups.
FAQ
Is GPT-6 Sol an official model now?
Yes. OpenAI released GPT-6 Sol and GPT-6 Luna on 2026-09-22; the official id is gpt-6-sol. Before that, the earliest source this page can trace was a community feature-request thread from 2026-09-11.
How much does GPT-6 Sol cost?
Official price: input $2, cached input $0.20, output $10 (per 1M tokens). Requests with more than 272K input tokens are billed at 2x input and cache rates and 1.5x output for the full request.
How does it compare with gpt-5.6-sol?
gpt-5.6-sol is officially $4 / $20. Calculated from the official price table, gpt-6-sol's input and output prices are both exactly half of that. gpt-5.6-sol is still available on QCode.
What about GPT-6 Luna and Terra?
GPT-6 Luna was released the same day as Sol, official id gpt-6-luna, $0.10 / $0.50; its official model page gives 1,050,000 context, 922,000 max input and 128,000 max output (checked 2026-09-29). Terra is not in the announcement's title; on QCode, gpt-5.6-terra remains available.
Did OpenAI publish benchmark scores?
OpenAI's 2026-09-22 launch post carries self-reported benchmarks; this page transcribes none of them and draws no conclusion about whether Sol or Luna is more capable. OpenAI describes the two as striking different balances of capability and cost.
How fresh is this?
Specs and prices come from archives of OpenAI's model page (2026-09-22) and pricing page (2026-09-23); the QCode model list reflects 2026-09-24. Defer to the official model and pricing pages and QCode's /models.
Sources
OpenAI's official news RSS (two posts on 2026-09-22), the GPT-6 Sol model page archived 2026-09-22, the official pricing page archived 2026-09-12 and 2026-09-23, and the OpenAI developer community thread. Defer to the official model and pricing pages.
Use gpt-6-sol directly
One QCode key covers gpt-6-sol, gpt-6-luna and gpt-6-astra — compare by switching the model id.
Related
Sol vs Terra vs Luna
What actually differs inside the official 5.6 family.
GPT-6 Astra quota burn
Real usage shape if you want generation 6.
GPT-5.5 retirement
A separate official event on 2026-10-14.
This page keeps officially confirmed facts apart from our own arithmetic: specs, prices and dates are quoted from official sources, and comparisons such as "half" are calculated from the official price table. It uses no benchmark scores and states no parameter the official material does not give.