GPT-5.6 Luna

OpenAI's cheap, fast 5.6 variant: 147 tokens a second at $0.20 per million in.

Official page ↗by OpenAI

Intelligence

AA Intelligence Index52

independent benchmark score from Artificial Analysis, not Timace's own judgment

Best for

✓ High-volume work where throughput matters more than the last few points of quality

Specifications

Developer: OpenAI
confidence 88%
artificialanalysis.ai
Verified Aug 20, 2026 · curated
Model type: Language
confidence 88%
artificialanalysis.ai
Verified Aug 20, 2026 · curated
Availability: Current
confidence 88%
artificialanalysis.ai
Verified Aug 20, 2026 · curated
AA Intelligence Index: 52
confidence 88%
artificialanalysis.ai
Verified Aug 20, 2026 · curated
Output speed (tok/s): 147 tok/s
confidence 88%
artificialanalysis.ai
Verified Aug 20, 2026 · curated
Latency to first token (s): 181.21s
confidence 88%
artificialanalysis.ai
Verified Aug 20, 2026 · curated
Input $/MTok: $0.20 / MTok
confidence 88%
benchlm.ai
Verified Aug 20, 2026 · curated
Output $/MTok: $1.20 / MTok
confidence 88%
benchlm.ai
Verified Aug 20, 2026 · curated
Pricing model: Usage based
confidence 88%
artificialanalysis.ai
Verified Aug 20, 2026 · curated
Context window (tokens): 1.05M tokens
confidence 88%
benchlm.ai
Verified Aug 20, 2026 · curated
Accepts: text, image, code
confidence 82%
artificialanalysis.ai
Verified Aug 20, 2026 · curated
Produces: text, code
confidence 88%
artificialanalysis.ai
Verified Aug 20, 2026 · curated
Extended reasoning: Yes
confidence 88%
artificialanalysis.ai
Verified Aug 20, 2026 · curated
Tool / function calling: Yes
confidence 88%
artificialanalysis.ai
Verified Aug 20, 2026 · curated
API available: Yes
confidence 88%
artificialanalysis.ai
Verified Aug 20, 2026 · curated
Open source: No
confidence 88%
artificialanalysis.ai
Verified Aug 20, 2026 · curated

Compare against

Side-by-side →

AI model

Amazon Polly NeuralThe default inside AWS: $16 per million characters and no new vendor to onboard.

AI model

Claude Fable 5Anthropic's most capable widely released model, for demanding reasoning and long-horizon agentic work.

AI model

Claude Haiku 4.5Near-frontier quality at commodity prices and high speed.

AI model

Claude Mythos 5Anthropic's limited-availability frontier model, Fable-tier pricing on a new tokenizer.

Specifications

Developer: OpenAI
confidence 88%
artificialanalysis.ai
Verified Aug 20, 2026 · curated
Model type: Language
confidence 88%
artificialanalysis.ai
Verified Aug 20, 2026 · curated
Availability: Current
confidence 88%
artificialanalysis.ai
Verified Aug 20, 2026 · curated
AA Intelligence Index: 52
confidence 88%
artificialanalysis.ai
Verified Aug 20, 2026 · curated
Output speed (tok/s): 147 tok/s
confidence 88%
artificialanalysis.ai
Verified Aug 20, 2026 · curated
Latency to first token (s): 181.21s
confidence 88%
artificialanalysis.ai
Verified Aug 20, 2026 · curated
Input $/MTok: $0.20 / MTok
confidence 88%
benchlm.ai
Verified Aug 20, 2026 · curated
Output $/MTok: $1.20 / MTok
confidence 88%
benchlm.ai
Verified Aug 20, 2026 · curated
Pricing model: Usage based
confidence 88%
artificialanalysis.ai
Verified Aug 20, 2026 · curated
Context window (tokens): 1.05M tokens
confidence 88%
benchlm.ai
Verified Aug 20, 2026 · curated
Accepts: text, image, code
confidence 82%
artificialanalysis.ai
Verified Aug 20, 2026 · curated
Produces: text, code
confidence 88%
artificialanalysis.ai
Verified Aug 20, 2026 · curated
Extended reasoning: Yes
confidence 88%
artificialanalysis.ai
Verified Aug 20, 2026 · curated
Tool / function calling: Yes
confidence 88%
artificialanalysis.ai
Verified Aug 20, 2026 · curated
API available: Yes
confidence 88%
artificialanalysis.ai
Verified Aug 20, 2026 · curated
Open source: No
confidence 88%
artificialanalysis.ai
Verified Aug 20, 2026 · curated