GPT-6 Luna

Sep 2026
GPT-6 Luna
Compare in
on
$0.029per task?
$0.10 / $0.50per 1M in / outcache $0.01
117 tok/sfirst token in 13 s?
1.05Mcontext

Rankingat high effort?

Full ranking →
1Muse Wonder Full32.92GPT-5.6 Sol32.33Opus 5.531.73GPT-6 Astra31.73GPT-6.1 Sol31.7
···
41Agnes 2.5 Pro Beta15.742GPT-6 Luna15.442Grok 4.515.4

Strong at

A
Reasoning#72 of 334Long context#52 of 273
B
Coding#44 of 132Agents & work tasks#44 of 175Knowledge#84 of 318Vision#35 of 124

Alternatives?

ModelScoreCostPer taskWhat changes
CheaperAgnes 2.5 Pro Betasame-25%$0.10 / $0.30Context: 1.05M → 1MBetterClaude Haiku 5.5high+20%+174%$0.079Context: 1.05M → 1M+6OpenMiMo v2.6 Prothinking+72%+359%$0.13Context: 1.05M → 1M+4Lower effortSame modelmedium-31%-40%$0.017BetterGPT-5.6 Lunaxhigh+33%+194%$0.085

History

2 versions
19.52020.5Release dateCritPt5.66
19.52020.5Release dateCritPt5.66

Benchmarks

in Chat & Knowledge
AA Intelligence Index 4.3.232.9#49
AA Omniscience Index-5.5#84
AA Omniscience · Accuracy42.8#40
AA Omniscience · Hallucination Rate84.4#197
Humanity's Last Exam32.9#72
AA-LCR (long context) 1.179.3#52
CritPt (physics)15.4#42
GDP.pdf · AA all-pass18.2#29
ARC-AGI-1 · public evaluation79.8#35
ARC-AGI-1 · semi-private70.3#39
ARC-AGI-2 · public evaluation31.9#34
ARC-AGI-2 · semi-private31.4#37
ARC-AGI-3 · semi-private0.2#17
HealthBench Professional length-adjustedlab60.8
HealthBench length-adjustedlab54.5
HealthBench Hard length-adjustedlab31.4
HealthBench Consensus length-adjustedlab95.9

Watch out 4

Function calling only with none · Chat Completions›

Chat Completions function calling requires reasoning_effort none; use Responses for reasoning with tools. Source ↗

More than 272K input tokens›

Prompts above 272K input tokens charge 2× input and cache rates and 1.5× output rates for the entire request. Source ↗

Remove unsupported sampling parameters›

When reasoning effort is not none, remove temperature, top_p and top_logprobs, plus Chat Completions logprobs or Responses output logprobs inclusion. Source ↗

Cache writes billed at 1.25x›

Cache writes cost 1.25× the uncached input token rate. Source ↗

FAQ

How much does GPT-6 Luna cost?›

$0.10 per 1M input tokens and $0.50 per 1M output tokens; cached input $0.01 per 1M. A typical benchmark task costs about $0.029 at high effort.

Is GPT-6 Luna free to use?›

There is no free API; the cheapest paid price is $0.10 per 1M input tokens.

How good is GPT-6 Luna, and what is it best at?›

#72 of 334 in Chat & Knowledge, #44 of 132 in Coding, #38 of 133 in Agents & Tool Use, #56 of 195 in Writing & Language, #35 of 124 in Vision & Documents. Strongest at: Reasoning, Long context.

Is GPT-6 Luna better than GPT-4.1?›

On CritPt (physics), GPT-6 Luna scores 19.4 vs 0 — higher.

How can I use GPT-6 Luna?›

Through the API of Amazon Bedrock, Azure, OpenAI.

What is the context window of GPT-6 Luna?›

1.05M tokens, with up to 128K output tokens.

When was GPT-6 Luna released? Is there a newer version?›

Released 2026-09-22. It is the current version in its line.

Is GPT-6 Luna open source?›

No, GPT-6 Luna is a closed model available through an API. The closest open-weights option here is MiMo v2.6 Pro.

Can GPT-6 Luna read images?›

It accepts text and images, PDF files as input.

What is GPT-6 Luna?›

GPT-6 Luna is a closed language model from OpenAI, released Sep 2026, the successor to GPT-4.1.

What are the best alternatives to GPT-6 Luna?›

Agnes 2.5 Pro Beta (cheaper), Claude Haiku 5.5 (better), MiMo v2.6 Pro (open weights).