HomeAI models
OpenAI: GPT-5.6 Luna Pro
OpenAI
•
gpt-5.6-luna-pro
•
Released 
July 2026

OpenAI
gpt-5.6-luna-pro

The pro-reasoning configuration of OpenAI's cheapest GPT-5.6 tier, buying frontier-adjacent intelligence at high-volume prices.

Modality:
Text
Image
PDF
model ID
openai/gpt-5.6-luna-pro
Start for free
arrow black icon
View API Docs
arrow white icon

Output Speed *

0.00
tok/s

Intelligence Index *

0
/ 100

Context Window *

1050000
tokens

Input price

1.2
Anytoken

Output price

7.2
Anytoken
GPT-5.6 Luna Pro: Deep Reasoning at the Family's Lowest Price Tier GPT-5.6 Luna Pro is OpenAI's pro-reasoning configuration of GPT-5.6 Luna, the fastest and most cost-efficient tier in the GPT-5.6 family below Terra and the Sol flagship. It is the same underlying Luna model served with reasoning.mode set to pro for higher-quality responses on complex tasks. It pairs extended chain-of-thought reasoning, a 1.05M-token context window, and text-plus-image input with per-token pricing well below the flagship. The workload that benefits most is high-volume reasoning: classification, extraction, and lightweight agentic steps where you need real reasoning quality without paying flagship rates. Access GPT-5.6 Luna Pro through the AnyAPI.ai unified API

Performance

Frontier-Adjacent Reasoning Without Frontier Pricing

Luna Pro's strength is extracting strong reasoning quality from the cheapest GPT-5.6 tier. On Artificial Analysis, GPT-5.6 Luna at max effort scores 52 on the Intelligence Index, well above the ~18 median for similarly priced reasoning models, while generating output around 126 tokens/second. Pro mode is a separate lever from reasoning effort: it increases model work for higher-quality responses on complex tasks, at the cost of more latency and tokens. For production, that means Luna Pro can handle genuinely hard reasoning steps at high-volume economics, but you must budget for the additional tokens pro mode consumes.

Benchmarks

Where Luna Sits on the Intelligence-vs-Cost Frontier

Artificial Analysis places GPT-5.6 Luna consistently on the Pareto frontier of intelligence versus cost per task. At max effort Luna scores 52 on the Intelligence Index; high effort scores 47 and medium 39, giving developers a tunable quality dial. Independent analysis found Luna at max buys roughly 86% of Sol max's measured intelligence for a small fraction of the per-task cost, and that Luna (max) matches or exceeds GLM-5.2 (max) and Gemini 3.5 Flash at lower cost. Note that these figures measure reasoning effort levels; pro mode itself is not represented in the standard 15-setting benchmark comparison.

Output Speed

*
0.00
tok/s

Intelligence Index

*
0
/ 100

MMLU *

Broad world knowledge and problem-solving
0
%

GPQA *

PhD-level scientific reasoning across physics, biology, chemistry.
0
%

HLE *

Adherence to multi-step structured instructions.
0
%

LiveCodeBench *

Tool-calling reliability in long agentic loops.
0
%

Technical Specifications

What the model supports

Luna Pro accepts text and image input plus files such as PDFs, and returns text only. It exposes a 1.05M-token context window with up to 128,000 completion tokens per request, a February 2026 knowledge cutoff, and reasoning.mode set to pro on top of Luna's effort settings (none, low, medium, high, xhigh, max). Function calling and JSON-schema structured outputs are supported. The two production-critical limits: output caps at 128K regardless of the huge input window, and prompts above 272K input tokens are repriced into a long-context tier, so large single-pass requests carry a cost penalty.
Verified Specifications — 
gpt-5.6-luna-pro
*
Input modalities
Text
Image
PDF
output modalities
Text
Context window
1050000
 tokens
Maximum output tokens
128000
Reasoning
Yes
Knowledge cutoff
July 2026
Pricing (standard)
1.2
 AnyTokens in
 / 
7.2
 AnyTokens out

Limitations & Trade-offs

Where gpt-5.6-luna-pro falls short

1
Long-context repricing penalty. Luna Pro exposes a 1.05M-token window, but prompts above 272K input tokens are billed at 2x input and 1.5x output for the full request. A small context increase can push an entire request into the long-context tier, so large single-pass document jobs lose the cheap-tier economics. For high-volume long-document work, selective retrieval usually beats stuffing the full window.
2
Pro mode inflates token usage and latency. reasoning.mode: pro is separate from reasoning.effort and increases model work, latency, and token consumption to improve quality on complex tasks. Because output is billed per token, indiscriminate use of pro mode across every request can erase the cost advantage of the Luna tier. It should be applied selectively to genuinely hard requests, not set globally.
3
High time to first token under heavy reasoning. Independent measurement of Luna at max effort showed a very high time to first token compared to non-reasoning peers, and deeper reasoning passes increase TTFT generally. For latency-sensitive, user-facing interactions, streaming helps perceived latency, but Luna Pro is a poor fit where the first token must appear near-instantly. Use lower effort or a non-reasoning model for those paths.
4
Text-only output and no disclosed size. Luna Pro accepts text and image input but returns text only — it does not generate images, audio, or video. OpenAI has not disclosed model size or parameter count. Teams needing multimodal generation must pair Luna Pro with a separate generation model, and cannot self-host or reason about capacity from a published parameter count.

Best-Fit Workloads

Where this model earns its place

01

High-volume structured extraction

‍
Luna Pro supports JSON-schema structured outputs and function calling, making it well suited to tagging, metadata generation, entity extraction, and form processing at scale. The pro-reasoning configuration adds accuracy on ambiguous inputs that a non-reasoning model would misclassify, while the low per-token rate keeps large batch jobs affordable. Watch tool-call and output-token economics, which can dominate an otherwise cheap Luna call.

02

Cost-sensitive reasoning agents
‍

For lightweight agentic workflows that need real reasoning between tool calls, Luna Pro delivers frontier-adjacent intelligence — Luna at max scored 52 on the Intelligence Index — at high-volume pricing. It supports tool_choice function calling and the Responses API. Route by uncertainty and consequence: use higher effort or pro mode only for hard steps, and escalate the rare highest-consequence decisions to Sol.

03

Long-document analysis with retrieval

‍
The 1.05M-token window lets Luna Pro read long reports, transcripts, or policy collections in a single pass with reasoning applied. In production, pair it with selective retrieval rather than always filling the window: retrieval improves relevance, cuts latency, and avoids the higher rates applied above the 272K-token threshold, keeping the cost-efficient tier genuinely cost-efficient.

04

Batch classification and moderation

‍
Independent benchmarks describe Luna as fast and highly concise, making it a strong fit for large-scale classification, routing, and content-moderation pipelines. Pro mode improves judgment on edge cases where a cheap non-reasoning model would fail, while the tier's economics support running the pipeline across very high request volumes. Keep effort low on the easy majority and reserve pro mode for borderline items.

Pricing in anytokens via AnyAPI
Input
1.2
₳
Output
7.2
₳
Cache write
1.5
₳
Cache read
0.12
₳
Get API access
arrow black icon
Read Docs

on this page

PerformanceBenchmarksTechnical SpecificationsLimitations & Trade-offsBest-Fit WorkloadsIntegrationFAQ

Integration

Access gpt-5.6-luna-pro via AnyAPI.ai

Access gpt-5.6-luna-pro through AnyAPI.ai using a unified API built for multi-model AI applications. Integrate gpt-5.6-luna-pro without maintaining a separate provider-specific connection, and keep the flexibility to test, switch, or combine models as your application requirements evolve.

01

One API integration

Access gpt-5.6-luna-pro and other AI models through the same API workflow instead of maintaining separate integrations for every provider.

02

Easy model switching

Test gpt-5.6-luna-pro against alternative models or switch models as your performance, capability, or cost requirements change without rebuilding your application around another provider API.

03

Flexible for production

Use gpt-5.6-luna-pro from experimentation through production while keeping your AI stack flexible as workloads, traffic, and model requirements evolve.

04

Multi-model applications

Use gpt-5.6-luna-pro for the workloads where it performs best and combine it with other models for tasks that require different capabilities, performance, or efficiency.

Start free
arrow black icon
Read Docs

Frequently Asked Questions

Answers to common questions about integrating and using this AI model via AnyAPI.ai

GPT-5.6 Luna Pro is OpenAI's pro-reasoning configuration of GPT-5.6 Luna, the fastest and most cost-efficient tier in the GPT-5.6 family. It is the same underlying Luna model served with reasoning.mode set to pro for higher-quality responses on complex tasks, using extended chain-of-thought reasoning before answering.

GPT-5.6 Luna Pro has a 1,050,000-token context window and can produce up to 128,000 output tokens per request. Note that prompts exceeding 272,000 input tokens are repriced into a long-context tier at 2x input and 1.5x output for the full request.

reasoning.mode: pro is separate from reasoning.effort. Effort (none, low, medium, high, xhigh, max) controls how hard the base model works; pro mode further increases model work for higher-quality responses on complex tasks, adding latency and token usage. Pro mode is not represented in the standard Artificial Analysis effort-based benchmark comparison.

Yes. GPT-5.6 Luna Pro accepts text and image input plus files such as PDFs, and returns text output. It supports function calling via tools and tool_choice, and structured outputs through a JSON schema in response_format. It does not generate images, audio, or video.

It is suited to cost-sensitive, high-volume reasoning and lightweight agentic workflows where you tune effort per task. For the most demanding agentic coding, long-horizon autonomy, or cybersecurity work, OpenAI positions the Sol flagship as the state-of-the-art choice; Luna Pro trades peak capability for far lower per-token cost.

* Benchmark data source: Artificial Analysis artificialanalysis.ai