Provider
OpenAI
Base model family
Model intelligence profile
GPT-5.6 Luna benchmark results and model details
This family profile brings together 9 published benchmark records for OpenAI GPT-5.6 Luna. Every result keeps its original benchmark, configuration, scale, and source; unrelated scores are never averaged.
Published benchmark evidence available · Business Skills V3 not yet authorized
Release date
Not yet verified
Exact-family source only
Published benchmarks
9
Native scales kept separate
Artificial Analysis
No exact match
As of 2026-07-16
Benchmark-level evidence
Published GPT-5.6 Luna benchmark results
These are individual benchmarks, not collection rollups. Scores remain on their original scales, and agent or harness results stay labelled as system configurations.
Spring Prompt benchmark
BulletBench
2 configurations
60-second bullet Elo
Chess decision quality under a shared 60-second game clock and a 10+1 per-move clock.
Ladder Elo: anchored to a Stockfish skill ladder (random mover = 400). Internally consistent ordering, not FIDE-calibrated.
2 published configurations
GPT-5.6 Luna (low)
147
- Bullet 95% interval
- 0–266
- 10+1 Elo
- 0
- Bullet games
- 96
- Bullet median move time
- 3.65s
- Average bullet game cost
- $0.0365
GPT-5.6 Luna (medium)
0
- Bullet 95% interval
- 0–83
- 10+1 Elo
- 0
- Bullet games
- 96
- Bullet median move time
- 3.90s
- Average bullet game cost
- $0.0372
Spring Prompt benchmark
ROASBench
4 configurations
Average ROASBench score
A 12-month performance-marketing simulation scored on business outcomes, planning, behavior, and persona fit.
4 published configurations
OpenAI: GPT-5.6 Luna
27.32
- Average contribution profit
- $-17,236
- Average ROAS
- 1.36×
- Completed runs
- 3
- Score variability
- ±0.04
OpenAI: GPT-5.6 Luna · High
35.34
- Average contribution profit
- $238,356
- Average ROAS
- 1.68×
- Completed runs
- 3
- Score variability
- ±1.04
OpenAI: GPT-5.6 Luna · Max
34.71
- Average contribution profit
- $246,328
- Average ROAS
- 1.69×
- Completed runs
- 3
- Score variability
- ±1.34
OpenAI: GPT-5.6 Luna · Medium
26.48
- Average contribution profit
- $39,991
- Average ROAS
- 1.44×
- Completed runs
- 3
- Score variability
- ±1.46
Source: Spring Prompt
Recorded 25 Jul 2026
Published ROASBench cache · 12-month simulation
Source record →
source-native benchmark
Artificial Analysis Intelligence Index
46
A composite index of language-model performance across agentic work, coding, scientific reasoning, knowledge, and long-context reasoning.
1 published configuration
GPT-5.6 Luna (high)
46
- Status
- Measured
- Recorded
- 30 Jul 2026
Source: Artificial Analysis
Recorded 30 Jul 2026
Intelligence Index v4.1 · 2026-07-30
Source record →
source-native benchmark
Coding Agent Configurations
6 configurations
Coding Agent Index v1.2 point score
Artificial Analysis Coding Agent Index v1.2 point scores and task-specific operational measurements for exact agent and model configurations.
These are exact agent-plus-model configurations, not bare-model scores.
6 published configurations
Codex - GPT-5.6 Luna (max)
53.8
- Mean cost per task
- $1.57
- Mean wall time
- 8.0 min
Codex - GPT-5.6 Luna (xhigh)
50.5
- Mean cost per task
- $1.26
- Mean wall time
- 6.6 min
Codex - GPT-5.6 Luna (high)
47.6
- Mean cost per task
- $0.96
- Mean wall time
- 5.7 min
Codex - GPT-5.6 Luna (medium)
38.4
- Mean cost per task
- $0.47
- Mean wall time
- 3.4 min
Codex - GPT-5.6 Luna (low)
23.4
- Mean cost per task
- $0.21
- Mean wall time
- 1.9 min
Codex - GPT-5.6 Luna (none)
17.9
- Mean cost per task
- $0.35
- Mean wall time
- 2.5 min
source-native benchmark
GDPval-AA v2
48%
Artificial Analysis's agentic evaluation of economically valuable, real-world work tasks based on the GDPval dataset.
1 published configuration
GPT-5.6 Luna (high)
48%
- Status
- Measured
- Recorded
- 30 Jul 2026
Source: Artificial Analysis
Recorded 30 Jul 2026
Independent evaluation · Public leaderboard snapshot · 2026-07-30
Source record →
source-native benchmark
GPQA Diamond
89%
The most challenging subset of Graduate-Level Google-Proof Q&A, focused on scientific reasoning.
1 published configuration
GPT-5.6 Luna (high)
89%
- Status
- Measured
- Recorded
- 30 Jul 2026
Source: Artificial Analysis
Recorded 30 Jul 2026
Independent evaluation · Public leaderboard snapshot · 2026-07-30
Source record →
source-native benchmark
Humanity's Last Exam
32%
A broad expert-level benchmark of difficult academic reasoning and knowledge questions.
1 published configuration
GPT-5.6 Luna (high)
32%
- Status
- Measured
- Recorded
- 30 Jul 2026
Source: Artificial Analysis
Recorded 30 Jul 2026
Independent evaluation · Public leaderboard snapshot · 2026-07-30
Source record →
source-native benchmark
MMMU-Pro
78%
A multimodal academic reasoning benchmark designed to reduce shortcuts and guessing across many disciplines.
1 published configuration
GPT-5.6 Luna (high)
78%
- Status
- Measured
- Recorded
- 30 Jul 2026
Source: Artificial Analysis
Recorded 30 Jul 2026
Independent evaluation · Public leaderboard snapshot · 2026-07-30
Source record →
source-native benchmark
Terminal-Bench v2.1
70%
A terminal-based agent benchmark covering software engineering, system administration, data processing, model training, and security tasks.
1 published configuration
GPT-5.6 Luna (high)
70%
- Status
- Measured
- Recorded
- 30 Jul 2026
Source: Artificial Analysis
Recorded 30 Jul 2026
Independent evaluation · Public leaderboard snapshot · 2026-07-30
Source record →
Independent operational reference
Artificial Analysis details
No exact, reviewed Artificial Analysis operational record is attached to this family. We do not inherit price, release, or throughput values from a nearby model name.
Early-user signal
Initial community opinions
No editor-reviewed community-opinions paragraph is active for this model. An active paragraph requires at least two retained Reddit discussions from the minimum 14-day launch window—implemented exactly as release date through day 14, end-exclusive—plus paraphrase-only review and a sealed publication artifact. Community reports stay separate from every benchmark score and model comparison.
Independent early-test signal
Early technical field tests
No independent field-test write-up has cleared review for this model yet — insufficient independent tests.
Business Skills V3 · proposed
How Spring Prompt plans to test GPT-5.6 Luna
The setup below is proposed and may change until preflight and execution approval are complete. Existing benchmark evidence above does not authorize or stand in for a Business Skills V3 result.
- Proposed reasoning
- High reasoning
- Provider revision
- Exact provider revision will be resolved and frozen only after preflight and execution approval
- Tools and service
- No tools · Provider-default service tier
- Planning configuration
- gpt-5.6-luna-high
- Generation controls
- Provider-managed reasoning; no temperature override proposed
Future first-party coverage
15 proposed Business Skills V3 task areas
These links describe evaluation contracts, not published GPT-5.6 Luna results.
Professional Knowledge Work & StrategyBusiness Workflow AutomationCustomer Support & Service ResolutionSales & MarketingLegal Document WorkFinance, Accounting & SpreadsheetsData Analysis & SQLResearch & Evidence SynthesisBusiness Writing & EmailDocument Comprehension & ExtractionSummarization & Grounded KnowledgeSoftware Engineering & CodingFrontend, UI & Web AppsTool Calling & Structured OutputPresentations, Charts & Visual Communication
Model comparisons
Compare GPT-5.6 Luna side by side
Editorially reviewed comparisons appear first. Every page matches only benchmark records with the same reviewed protocol key and does not manufacture an overall winner.
Publication safeguards
What must pass before a V3 result appears
- 1First-party task-local comparisons and eligible external evidence must both be present.
- 2Model and provider configuration identity must match the reviewed release exactly.
- 3Coverage, reliability, judge diagnostics, and sealed stability checks must pass.
- 4Uncertainty and missing evidence remain visible when results are published.
Stable family URL
Evidence can grow without changing the page
New reviewed benchmark snapshots, operational facts, community themes, and early field-test syntheses can be added here while the canonical model-family identity remains fixed.
More from OpenAI