Data as of November 6, 2026 · How the score is built
Released Sep 22, 20261.05M contextOpenAI GPT-6 Luna model documentation
GPT-6 Luna
Decision readingGPT-6 Luna scores 62.6 out of 100 and ranks #43 of 212. This profile shows 26 source-displayable benchmark rows; its strongest eligible category is Reasoning at #22. API pricing is $0.1 input and $0.5 output per million tokens, with cached input at $0.01.
Released Sep 22, 2026 — see all recent releases
Decision snapshot
Each value carries a field reference instead of floating alone. Markers compare this model with the current ranked and priced catalog; they are not absolute quality thresholds.
Capability
62.6/100
field median 50.1#43 of 212 ranked models
Public
#43of 212
Verified #28 of 77
Price
$0.10input / $0.50 output
input median $0.60cached $0.010 · blended $0.30
Speed
Not measured
field median 92 tok/sTime to first token not measured
Context
1.05Mtokens
field median 256,000Maximum output length is tracked separately
Strongest published evidence
Reasoning ranks #22. Particularly strong for complex reasoning, multi-step problem solving, and analytical tasks.
Validate before choosing
26 published rows leave some tracked benchmark slots empty. Independent runtime speed has not been measured.
Source-linked · 26 displayable benchmark rows
GPT-6 Luna API pricing
GPT-6 Luna costs $0.10 per million input tokens and $0.50 per million output tokens on the OpenAI API. Cached input is $0.01 per million.
| Token type | Price per 1M tokens |
|---|---|
| Input | $0.10 |
| Cached input | $0.01 |
| Output | $0.50 |
At these rates, a request with 10,000 input tokens and 1,000 output tokens costs $0.0015 without caching.
On blended price (the average of the input and output rates), 1 of the 112 other ranked models with a published API price costs the same or less and scores higher. See where it sits on the chart
Pricing notes: OpenAI's GPT-6 Luna model page lists $0.10 input / $0.01 cached input / $0.125 cache writes / $0.50 output per million tokens. Prompts above 272K input tokens are charged at 2x input and cache rates and 1.5x output for the full request; regional processing adds 10% where available; Batch and Flex are 50% of Standard and Fast mode is 2x.
Rates from OpenAI GPT-6 Luna model documentation. API model ID: gpt-6-luna. Price registry updated November 6, 2026.
Category score record
Scores and ranks appear only where published evidence can be displayed. The table keeps the score, weight, cohort, and evidence state together.
| Category | Score | Rank | Percentile | Weight | Benchmarks | Evidence |
|---|---|---|---|---|---|---|
| Agentic5.2 points below GPT-5.6 LunaRank #45 of 131Percentile 66thWeight 22%5 benchmarksVerified | 53.5 | 5 benchmarks | Verified | |||
| Coding10.1 points below GPT-5.6 LunaRank #42 of 157Percentile 74thWeight 20%3 benchmarksVerified | 51.8 | 3 benchmarks | Verified | |||
| ReasoningRank #22 of 25Percentile 13thWeight 17%5 benchmarksVerified | 57.5 | 5 benchmarks | Verified | |||
| MultimodalRank Not rankedWeight 12%3 benchmarksVerified | 81.9 | 3 benchmarks | Verified | |||
| KnowledgeRank #40 of 178Percentile 78thWeight 12%5 benchmarksVerified | 63.3 | 5 benchmarks | Verified | |||
| MultilingualWeight 7%0 benchmarksNot measured | Not measured | 0 benchmarks | Not measured | |||
| Inst. FollowingWeight 5%0 benchmarksNot measured | Not measured | 0 benchmarks | Not measured | |||
| MathRank Not rankedWeight 5%2 benchmarksVerified | 80.2 | 2 benchmarks | Verified |
An earlier release can outscore a newer one in a single category. The note under that category names the release and the gap.
26 of 816 tracked benchmark slots have displayable evidence · bars run 0–100
Coverage detailsHow much of this is verified
Coverage is split by category so a strong number never hides a thin evidence base. Verified means the row is tied to a published source; provisional rows remain visible but separate.
- Agentic5/5 verified
- Coding3/3 verified
- Reasoning5/5 verified
- Multimodal3/3 verified
- Knowledge5/5 verified
- MultilingualNot measured
- Inst. FollowingNot measured
- Math2/2 verified
Capability shape
Each axis shows percentile within that category’s eligible cohort. The comparison outline is the median of the six nearest public-score peers; a collapsed vertex means the category is not rank-eligible.
GPT-6 Luna category percentile values
- Agentic66th percentile
- Coding74th percentile
- Reasoning13th percentile
- MultimodalNot eligible
- Knowledge78th percentile
- MultilingualNot eligible
- Instruction followingNot eligible
- MathNot eligible
The dashed outline is median of 6 nearest peers.
Eligible category ranks
- Agentic#45/131
- Coding#42/157
- Reasoning#22/25
- MultimodalNot ranked
- Knowledge#40/178
- MultilingualNot ranked
- Inst. FollowingNot ranked
- MathNot ranked
Benchmark ledger
Coding opens by default. The marker compares each value with the best source-verified result in the catalog; provisional leaders do not set the reference. Expand the remaining categories for every published row.
Coding3 rows
| Benchmark | Score | Versus best verified row | Gap | Weight | Evidence |
|---|---|---|---|---|---|
| DeepSWE | Score66.6% | Versus best verified row Best verified: Muse Wonder Full · 79.0% | Gap12.4 behind | Weight15% ref. weight | Provider exact |
| Bug Hunt Bench | Score18.3 fixes | Versus best verified row Best verified: Claude Sonnet 5.5 · 51.3 fixes | Gap33 behind | WeightDisplay only | Benchmark exact |
| FrontierCode 2.0 | Score21.9% | Versus best verified row Best verified: Claude Opus 5.5 · 45.5% | Gap23.6 behind | WeightDisplay only | Benchmark exact |
Agentic5 rows
| Benchmark | Score | Versus best verified row | Gap | Weight | Evidence |
|---|---|---|---|---|---|
| ExploitGym | Score11.6% | Versus best verified row Best verified: Gemini 4 Carbon · 53.5% | Gap41.9 behind | WeightDisplay only | Provider exact |
| Hermes IndexHermes Index: agentic capability and cost per task inside Hermes Agent | Score33.9% | Versus best verified row Best verified: Claude Opus 5.5 · 63.3% | Gap29.4 behind | WeightDisplay only | Benchmark exact |
| Hermes BenchHermes Bench: 150 agent tasks graded on the files and state the agent leaves behind | Score70.5% | Versus best verified row Best verified: Claude Opus 5.5 · 76.7% | Gap6.3 behind | WeightDisplay only | Benchmark exact |
| MazeBenchMazeBench: visual spatial reasoning in a 3D open world | Score0% | Versus best verified row Best verified: GPT-6 Astra · 23% | Gap23 behind | WeightDisplay only | Benchmark exact |
| OSWorld 2.1 OfflineOSWorld 2.1 offline subset (82 tasks) | Score48.9% | Versus best verified row Best verified: Claude Opus 5.5 · 87.2% | Gap38.3 behind | WeightDisplay only | Provider exact |
Reasoning5 rows
| Benchmark | Score | Versus best verified row | Gap | Weight | Evidence |
|---|---|---|---|---|---|
| ARC-AGI-2Abstraction and Reasoning Corpus for AGI v2 | Score59.3% | Versus best verified row Best verified: GPT-6 Astra · 95% | Gap35.7 behind | WeightWeighted 25% | Benchmark exact |
| ARC-AGI-3Abstraction and Reasoning Corpus for AGI v3 | Score0.1% | Versus best verified row Best verified: GPT-6 Astra · 62.7% | Gap62.6 behind | WeightWeighted 15% | Benchmark exact |
| BioBasicsBenchBioBasicsBench: everyday wet-lab science | Score37.6% | Versus best verified row Best verified: GPT-6 Astra · 67.0% | Gap29.4 behind | WeightDisplay only | Provider exact |
| ARC-AGI-1ARC-AGI-1 Semi-Private Evaluation | Score86.67% | Versus best verified row Best verified: Claude Fable 5 · 98.50% | Gap11.8 behind | WeightDisplay only | Benchmark exact |
| ARC-AGI-3 Provider AdapterARC-AGI-3 Semi-Private: Provider Adapter harness | Score0.59% | Versus best verified row Best verified: GPT-6 Astra · 98.55% | Gap98 behind | WeightDisplay only | Benchmark exact |
Multimodal3 rows
| Benchmark | Score | Versus best verified row | Gap | Weight | Evidence |
|---|---|---|---|---|---|
| ExtractBench ProExtractBench Pro: long-tail document extraction challenges (LlamaIndex) | Score0% | Versus best verified row Best verified: GPT-6 Astra · 58% | Gap58 behind | WeightDisplay only | |
| ExtractBench (LlamaIndex)ExtractBench: schema-guided enterprise document extraction (LlamaIndex leaderboard) | Score91.01% | Versus best verified row Best verified: GPT-6.1 Sol · 95.31% | Gap4.3 behind | WeightDisplay only | |
| ParseBenchParseBench: document parsing for AI agents (LlamaIndex leaderboard) | Score64.13% | Versus best verified row Best verified: Claude Opus 5.5 · 78.80% | Gap14.7 behind | WeightDisplay only |
Knowledge5 rows
| Benchmark | Score | Versus best verified row | Gap | Weight | Evidence |
|---|---|---|---|---|---|
| HealthBench (raw)HealthBench raw score | Score50.0% | Versus best verified row Best verified: Claude Sonnet 5.5 · 69.4% | Gap19.4 behind | WeightDisplay only | Provider exact |
| HealthBench (length-adjusted)HealthBench length-adjusted score | Score54.5% | Versus best verified row Best verified: Claude Sonnet 5.5 · 65.4% | Gap10.9 behind | WeightDisplay only | Provider exact |
| HealthBench Professional | Score60.8% | Versus best verified row Best verified: Claude Sonnet 5.5 · 69.2% | Gap8.4 behind | WeightDisplay only | Provider exact |
| HealthBench Professional (raw)HealthBench Professional raw score | Score61.2% | Versus best verified row Best verified: Claude Sonnet 5.5 · 77.1% | Gap15.9 behind | WeightDisplay only | Provider exact |
| HealthBench Hard | Score31.4% | Versus best verified row Best verified: Muse Spark · 42.8% | Gap11.4 behind | WeightDisplay only | Provider exact |
Math2 rows
| Benchmark | Score | Versus best verified row | Gap | Weight | Evidence |
|---|---|---|---|---|---|
| FrontierMath v2 (Tiers 1-3)FrontierMath v2 Tiers 1-3 | Score78.947% | Versus best verified row Best verified: GPT-6 Astra · 93.684% | Gap14.7 behind | WeightWeighted 30% | Benchmark exact |
| FrontierMath v2 (Tier 4)FrontierMath v2 Tier 4 | Score56.098% | Versus best verified row Best verified: GPT-6.1 Sol · 100.000% | Gap43.9 behind | WeightWeighted 10% | Benchmark exact |
External signals3 rows
| Benchmark | Score | Versus best verified row | Gap | Weight | Evidence |
|---|---|---|---|---|---|
| SEC-Bench Pro | Score34.2% | Versus best verified row Best verified: Gemini 4 Carbon · 88.2% | Gap54 behind | WeightDisplay only | Provider exact |
| ExploitBench Internal PortExploitBench — Internal Port (June–August 2026) | Score0.0% | Versus best verified row Best verified: GPT-6 Cyber · 42.0% | Gap42 behind | WeightDisplay only | Provider exact |
| ExploitGym HoneypotExploitGym Honeypot out-of-scope exploitation rate | Score0.00% | Versus best verified row Best verified: GPT-6 Astra · 0.00% | GapBest verified | WeightDisplay only | Provider exact |
What it costs to get this score
Published API price against the public score. The x-axis uses a log scale; the dashed path marks models that are not beaten by a cheaper, higher-scoring option. Price uses average of published input and output rates.
GPT-6 Luna · 62.6 score · $0.30 blended per million tokens
The chart opens on the current model. Scroll horizontally to inspect the full price axis.
Horizontal: blended price per million tokens, log scale · Vertical: public score
Lineage
The sequence follows explicit supersedes links. Each score is estimated for that model; a relative can inform a sparse estimate but never sets a floor, so a newer release can score below an earlier one. Scores and prices remain blank when the corresponding public row or first-party rate is unavailable.
Spec sheet
Each documented value carries its source. Missing fields stay visible as not sourced or not published, rather than disappearing from the page.
- API model ID
- gpt-6-lunaOpenAI GPT-6 Luna model documentation
- Context window
- 1.05MOpenAI GPT-6 Luna model documentation
- Maximum output
- 128,000 tokensOpenAI GPT-6 Luna model documentation
- Knowledge cutoff
- May 18, 2026OpenAI GPT-6 Luna model documentation
- Input modalities
- text, imageOpenAI GPT-6 Luna model documentation
- Output modalities
- textOpenAI GPT-6 Luna model documentation
- Parameters
- Not disclosed by the provider
- Availability
- OpenAI Responses API · OpenAI Chat Completions API · ChatGPT Work · CodexOpenAI GPT-6 Sol and Luna launch
Chat Completions supports function calling with GPT-6 Luna only when reasoning_effort is none. Use the Responses API for built-in tools or for function calling with reasoning on. OpenAI GPT-6 Luna model documentation, read Sep 25, 2026
- Cloud regions
- Not tracked yet
- Lifecycle
- activeOpenAI GPT-6 Luna model documentation
- API capabilities
- Streaming · function calling · structured outputs · web/file search · image generation · code interpreter · hosted shell · apply patch · skills · computer use · MCP · tool searchOpenAI GPT-6 Luna model documentation
- Prompt caching
- Published at $0.010 per million cached input tokensOpenAI GPT-6 Luna model documentation
- Self-host
- Weights are not published
- Rate limits
- Tier 1: 500 RPM / 500K TPM; Tier 5: 30K RPM / 180M TPM; Free unsupportedOpenAI GPT-6 Luna model documentation
How to read this profile
The visual layer above carries the decisions. These notes preserve the model, ranking, coverage, and family context behind the numbers.
GPT-6 Luna ranks #43 of 212 on the public leaderboard with a score of 62.59/100. Its source-verified position is #28 of 77.
GPT-6 Luna is a proprietary model with a 1.05M context window. It uses an explicit reasoning mode, which can improve complex problem solving while adding latency and token use.
Released September 22, 2026 in ChatGPT Work and Codex for Plus, Pro, Business, Enterprise, and Edu users, and in the OpenAI API as gpt-6-luna. OpenAI documents Responses API support for built-in tools and function calling; Chat Completions supports function calling when reasoning_effort is set to none.
OpenAI documents GPT-6 Luna as a distinct API model for focused, high-volume tasks, with configurable reasoning, text and image input, text output, and a 1,050,000-token context window. OpenAI publishes provider-run measurements, and ARC Prize provides verified ARC-AGI results.
GPT-6 Luna sits in the GPT-6 family with GPT-6 Astra, GPT-6.1 Sol, GPT-6.1 Terra, GPT-6 Sol, GPT-6 Cyber. Its explicit predecessor is GPT-5.6 Luna. 26 of 816 tracked benchmark slots currently have displayable evidence. Missing categories stay blank.
Its strongest eligible category is Reasoning at #22, while its lowest eligible position is Agentic at #45. particularly strong for complex reasoning, multi-step problem solving, and analytical tasks.
Last updated November 6, 2026. Runtime fields remain blank until a sourced snapshot exists.
Deployment options
Self-host and provider-specific paths stay separate from benchmark evidence so operating constraints are visible before a score becomes the whole decision.
Where to run GPT-6 Luna by API
OpenRouter is the one host we track that publishes a per-token price for it. Prices read from each host’s own pages on October 20, 2026.
| Host | Input /1M | Output /1M | Cached /1M | Context |
|---|---|---|---|---|
| OpenRouter | $0.10 | $0.50 | $0.01 | 1.05M |
Questions
How does GPT-6 Luna perform overall in AI benchmarks?
GPT-6 Luna ranks #43 out of 212 models on the public BenchAlign leaderboard, with a score of 62.59/100. Its evidence status is Supported, and this profile shows 26 source-displayable benchmark rows. The label describes evidence depth, not a provider quality claim; inspect category rows before choosing a workload.
Is GPT-6 Luna good for knowledge and understanding?
GPT-6 Luna ranks #40 out of 178 eligible models for knowledge and understanding, with a public category score of 63.3/100. Higher-ranked alternatives are available for workloads where this category decides the choice. Check the underlying rows before treating the aggregate as a workload guarantee.
Is GPT-6 Luna good for coding and programming?
GPT-6 Luna ranks #42 out of 157 eligible models for coding and programming, with a public category score of 51.8/100. Higher-ranked alternatives are available for workloads where this category decides the choice. Check the underlying rows before treating the aggregate as a workload guarantee.
Is GPT-6 Luna good for mathematics?
GPT-6 Luna has source-displayable benchmark coverage for mathematics, but the public category table does not assign it a rank there. The individual rows remain available for inspection. A missing category position means the evidence threshold was not met; it does not convert the model's unmeasured work into a zero.
Is GPT-6 Luna good for reasoning and logic?
GPT-6 Luna ranks #22 out of 25 eligible models for reasoning and logic, with a public category score of 57.5/100. Higher-ranked alternatives are available for workloads where this category decides the choice. Check the underlying rows before treating the aggregate as a workload guarantee.
Is GPT-6 Luna good for agentic tool use and computer tasks?
GPT-6 Luna ranks #45 out of 131 eligible models for agentic tool use and computer tasks, with a public category score of 53.5/100. Higher-ranked alternatives are available for workloads where this category decides the choice. Check the underlying rows before treating the aggregate as a workload guarantee.
Is GPT-6 Luna good for multimodal and grounded tasks?
GPT-6 Luna has source-displayable benchmark coverage for multimodal and grounded tasks, but the public category table does not assign it a rank there. The individual rows remain available for inspection. A missing category position means the evidence threshold was not met; it does not convert the model's unmeasured work into a zero.
Which sibling models are related to GPT-6 Luna?
GPT-6 Luna belongs to the GPT-6 family. Related tracked variants include GPT-6 Astra, GPT-6.1 Sol, GPT-6.1 Terra, plus 2 more. A sibling link indicates shared lineage or a documented configuration relationship; it does not mean the variants have identical pricing, context limits, benchmark evidence, or deployment behavior. Compare before switching.
Does GPT-6 Luna have full benchmark coverage on BenchLM?
No. GPT-6 Luna currently has 47 source-displayable rows. The profile exposes published, non-generated evidence and leaves missing categories blank until an exact evaluation is available. Coverage describes how much was measured; it is not a penalty added to an individual benchmark result.
What is the context window size of GPT-6 Luna?
GPT-6 Luna has a documented context window of 1.05M. That figure is the maximum combined prompt and retained-conversation space reported for this exact model; it is not the maximum output length. The profile keeps output limits separate because providers often publish those limits independently.