Skip to main content
The Human Bit

OpenAI · GPT-6

GPT-6 Luna

OpenAI's fastest, cheapest current model for focused, high-volume tasks such as summarising, extracting and lightweight coding, released 22 September 2026 as the successor to GPT-5.6 Luna. Free and Go users can try it in the ChatGPT desktop app; GPT-5.6 Luna remains available during the rollout.

Start here

What this helps you do
Decide whether GPT-6 Luna is the right model for a job you have, or one to leave alone.
What you leave with
A plain read of what GPT-6 Luna is for, what to avoid giving it, and the facts behind that.
What to check before trusting it
Released 22 September 2026, so real-world track record is minimal at review time.
Ask Bit which of your jobs suit GPT-6 Luna
The Human Bit position

A reasonable default for high-volume, low-stakes work once you have checked it against GPT-5.6 Luna on your own task; the price drop is real, but confirm quality holds before routing volume through it.

Use it for

  • Summarisation and extraction at volume
  • Focused, repetitive coding tasks
  • Quick everyday work where Sol's extra reasoning is not needed
Leave it alone when

Do not reach for it on genuinely hard multi-step reasoning or long-horizon agentic work; that is what Sol and Astra are for.

Effort without the jargon

Up to Max

OpenAI documents Luna's reasoning effort as going up to Max but not Ultra; use higher effort only for tasks that need more than a quick pass, and reach for Sol once that is not enough.

Advanced recordCost, limits, privacy and operations

This is the cheap, high-volume option; confirm it holds quality on your own summarisation or extraction task before routing real volume through it.

API cost
Officially disclosedOpenAI API price: $0.10 per million input tokens, $0.01 cached input and $0.50 per million output tokens.
Context window
Officially disclosed1,050,000 tokens
Maximum output
Officially disclosed128,000 tokens
Knowledge cutoff
Officially disclosed18 May 2026
Recorded inputs
Not enumerated in cited record
Recorded output
Text

Data handling is a product decision

Depends on surfaceRetention and training rules belong to the ChatGPT plan, API account and workspace, not the model name alone; the cited model page does not restate them.

Operational constraints

  • Released 22 September 2026, so real-world track record is minimal at review time.
  • Reasoning effort goes up to Max but not Ultra, per OpenAI's own product documentation.
  • GPT-5.6 Luna remains available during the rollout, so this is not a forced migration.
  • The cited model page does not enumerate input modalities beyond text; treat image input as unconfirmed until checked against a live request.

Record details

Evidence readiness
informational
Guidance confidence
moderate
Evidence revision
gpt-6-luna@v1

Unknown means the cited official record does not disclose a safe value. It is not an estimate. Prices are provider list prices in USD where stated and can change before this record's review date.

Keep the evidence separate

Canonical provider claim

OpenAI describes GPT-6 Luna as its most efficient model for focused, high-volume tasks including summarisation, extraction and focused coding, built on the advances behind GPT-6 Astra brought into a faster, more affordable model.

Independent observation

No sufficiently useful independent note has been added yet. That absence is not filled with our guess.

Readiness · informational

No Human Bit result is claimed; use this as evaluation guidance only.

Human Bit test

Not scheduled

No Human Bit result is claimed for this model yet.

Sources

Compare GPT-6 Luna with another route for a job →