Name: GPT-5.6 Luna vs GPT-6 Luna
Creator: Benchmark Registry

Shared benchmarks

Shared benchmarks for GPT-5.6 Luna and GPT-6 Luna
Benchmark	GPT-5.6 Luna	GPT-6 Luna
AutomationBench 1.0.6	9.1%1.0.6 · Task pass rate OpenAI 12.9%1.0.6 · Task pass rate OpenAI 1.8%1.0.6 · Task pass rate OpenAI 4.3%1.0.6 · Task pass rate OpenAI 17.0%1.0.6 · Task pass rate OpenAI	12.6%1.0.6 · Task pass rate OpenAI 14.5%1.0.6 · Task pass rate OpenAI 1.2%1.0.6 · Task pass rate OpenAI 9.4%1.0.6 · Task pass rate OpenAI 20.7%1.0.6 · Task pass rate OpenAI
Evaluation details for AutomationBench 1.0.6 The recorded benchmark version, metric, and evaluator sets match. Evaluation methodology is not recorded by the registry. GPT-5.6 Luna Score 9.1% Evaluator OpenAI Version AutomationBench 1.0.6 Metric Task pass rate (percent) Reasoning high Reported September 22, 2026 Source Evaluation source for GPT-5.6 Luna on AutomationBench 1.0.6 (opens in a new tab) Score 12.9% Evaluator OpenAI Version AutomationBench 1.0.6 Metric Task pass rate (percent) Reasoning xhigh Reported September 22, 2026 Source Evaluation source for GPT-5.6 Luna on AutomationBench 1.0.6 (opens in a new tab) Score 1.8% Evaluator OpenAI Version AutomationBench 1.0.6 Metric Task pass rate (percent) Reasoning low Reported September 22, 2026 Source Evaluation source for GPT-5.6 Luna on AutomationBench 1.0.6 (opens in a new tab) Score 4.3% Evaluator OpenAI Version AutomationBench 1.0.6 Metric Task pass rate (percent) Reasoning medium Reported September 22, 2026 Source Evaluation source for GPT-5.6 Luna on AutomationBench 1.0.6 (opens in a new tab) Score 17.0% Evaluator OpenAI Version AutomationBench 1.0.6 Metric Task pass rate (percent) Reasoning max Reported September 22, 2026 Source Evaluation source for GPT-5.6 Luna on AutomationBench 1.0.6 (opens in a new tab) GPT-6 Luna Score 12.6% Evaluator OpenAI Version AutomationBench 1.0.6 Metric Task pass rate (percent) Reasoning xhigh Reported September 22, 2026 Source Evaluation source for GPT-6 Luna on AutomationBench 1.0.6 (opens in a new tab) Score 14.5% Evaluator OpenAI Version AutomationBench 1.0.6 Metric Task pass rate (percent) Reasoning high Reported September 22, 2026 Source Evaluation source for GPT-6 Luna on AutomationBench 1.0.6 (opens in a new tab) Score 1.2% Evaluator OpenAI Version AutomationBench 1.0.6 Metric Task pass rate (percent) Reasoning low Reported September 22, 2026 Source Evaluation source for GPT-6 Luna on AutomationBench 1.0.6 (opens in a new tab) Score 9.4% Evaluator OpenAI Version AutomationBench 1.0.6 Metric Task pass rate (percent) Reasoning medium Reported September 22, 2026 Source Evaluation source for GPT-6 Luna on AutomationBench 1.0.6 (opens in a new tab) Score 20.7% Evaluator OpenAI Version AutomationBench 1.0.6 Metric Task pass rate (percent) Reasoning max Reported September 22, 2026 Source Evaluation source for GPT-6 Luna on AutomationBench 1.0.6 (opens in a new tab)
DeepSWE 1.1	9.3%1.1 · DeepSWE Pass@1 OpenAI 42.4%1.1 · DeepSWE Pass@1 OpenAI 62.2%1.1 · DeepSWE Pass@1 OpenAI 1.2%1.1 · DeepSWE Pass@1 OpenAI 56.2%1.1 · DeepSWE Pass@1 OpenAI 67.2%1.1 · DeepSWE Pass@1 OpenAI	66.6%1.1 · DeepSWE Pass@1 OpenAI 59.3%1.1 · DeepSWE Pass@1 OpenAI 2.4%1.1 · DeepSWE Pass@1 OpenAI 61.3%1.1 · DeepSWE Pass@1 OpenAI 44.5%1.1 · DeepSWE Pass@1 OpenAI
Evaluation details for DeepSWE 1.1 The recorded benchmark version, metric, and evaluator sets match. Evaluation methodology is not recorded by the registry. GPT-5.6 Luna Score 9.3% Evaluator OpenAI Version DeepSWE 1.1 Metric DeepSWE Pass@1 (percent) Reasoning medium Reported September 22, 2026 Source Evaluation source for GPT-5.6 Luna on DeepSWE 1.1 (opens in a new tab) Score 42.4% Evaluator OpenAI Version DeepSWE 1.1 Metric DeepSWE Pass@1 (percent) Reasoning high Reported September 22, 2026 Source Evaluation source for GPT-5.6 Luna on DeepSWE 1.1 (opens in a new tab) Score 62.2% Evaluator OpenAI Version DeepSWE 1.1 Metric DeepSWE Pass@1 (percent) Reasoning max Reported September 22, 2026 Source Evaluation source for GPT-5.6 Luna on DeepSWE 1.1 (opens in a new tab) Score 1.2% Evaluator OpenAI Version DeepSWE 1.1 Metric DeepSWE Pass@1 (percent) Reasoning low Reported September 22, 2026 Source Evaluation source for GPT-5.6 Luna on DeepSWE 1.1 (opens in a new tab) Score 56.2% Evaluator OpenAI Version DeepSWE 1.1 Metric DeepSWE Pass@1 (percent) Reasoning xhigh Reported September 22, 2026 Source Evaluation source for GPT-5.6 Luna on DeepSWE 1.1 (opens in a new tab) Score 67.2% Evaluator OpenAI Version DeepSWE 1.1 Metric DeepSWE Pass@1 (percent) Reasoning Not specified Reported July 9, 2026 Source Evaluation source for GPT-5.6 Luna on DeepSWE 1.1 (opens in a new tab) GPT-6 Luna Score 66.6% Evaluator OpenAI Version DeepSWE 1.1 Metric DeepSWE Pass@1 (percent) Reasoning max Reported September 22, 2026 Source Evaluation source for GPT-6 Luna on DeepSWE 1.1 (opens in a new tab) Score 59.3% Evaluator OpenAI Version DeepSWE 1.1 Metric DeepSWE Pass@1 (percent) Reasoning high Reported September 22, 2026 Source Evaluation source for GPT-6 Luna on DeepSWE 1.1 (opens in a new tab) Score 2.4% Evaluator OpenAI Version DeepSWE 1.1 Metric DeepSWE Pass@1 (percent) Reasoning low Reported September 22, 2026 Source Evaluation source for GPT-6 Luna on DeepSWE 1.1 (opens in a new tab) Score 61.3% Evaluator OpenAI Version DeepSWE 1.1 Metric DeepSWE Pass@1 (percent) Reasoning xhigh Reported September 22, 2026 Source Evaluation source for GPT-6 Luna on DeepSWE 1.1 (opens in a new tab) Score 44.5% Evaluator OpenAI Version DeepSWE 1.1 Metric DeepSWE Pass@1 (percent) Reasoning medium Reported September 22, 2026 Source Evaluation source for GPT-6 Luna on DeepSWE 1.1 (opens in a new tab)
OSWorld 2.0 v2026.08.08 Offline — Partial	52.7%2.0 v2026.08.08 Offline — Partial · OSWorld Partial Score OpenAI 35.9%2.0 v2026.08.08 Offline — Partial · OSWorld Partial Score OpenAI 22.7%2.0 v2026.08.08 Offline — Partial · OSWorld Partial Score OpenAI 48.0%2.0 v2026.08.08 Offline — Partial · OSWorld Partial Score OpenAI 12.0%2.0 v2026.08.08 Offline — Partial · OSWorld Partial Score OpenAI	46.7%2.0 v2026.08.08 Offline — Partial · OSWorld Partial Score OpenAI 31.5%2.0 v2026.08.08 Offline — Partial · OSWorld Partial Score OpenAI 52.7%2.0 v2026.08.08 Offline — Partial · OSWorld Partial Score OpenAI 41.4%2.0 v2026.08.08 Offline — Partial · OSWorld Partial Score OpenAI 8.3%2.0 v2026.08.08 Offline — Partial · OSWorld Partial Score OpenAI
Evaluation details for OSWorld 2.0 v2026.08.08 Offline — Partial The recorded benchmark version, metric, and evaluator sets match. Evaluation methodology is not recorded by the registry. GPT-5.6 Luna Score 52.7% Evaluator OpenAI Version OSWorld 2.0 v2026.08.08 Offline — Partial Metric OSWorld Partial Score (percent) Reasoning max Reported September 22, 2026 Source Evaluation source for GPT-5.6 Luna on OSWorld 2.0 v2026.08.08 Offline — Partial (opens in a new tab) Score 35.9% Evaluator OpenAI Version OSWorld 2.0 v2026.08.08 Offline — Partial Metric OSWorld Partial Score (percent) Reasoning high Reported September 22, 2026 Source Evaluation source for GPT-5.6 Luna on OSWorld 2.0 v2026.08.08 Offline — Partial (opens in a new tab) Score 22.7% Evaluator OpenAI Version OSWorld 2.0 v2026.08.08 Offline — Partial Metric OSWorld Partial Score (percent) Reasoning medium Reported September 22, 2026 Source Evaluation source for GPT-5.6 Luna on OSWorld 2.0 v2026.08.08 Offline — Partial (opens in a new tab) Score 48.0% Evaluator OpenAI Version OSWorld 2.0 v2026.08.08 Offline — Partial Metric OSWorld Partial Score (percent) Reasoning xhigh Reported September 22, 2026 Source Evaluation source for GPT-5.6 Luna on OSWorld 2.0 v2026.08.08 Offline — Partial (opens in a new tab) Score 12.0% Evaluator OpenAI Version OSWorld 2.0 v2026.08.08 Offline — Partial Metric OSWorld Partial Score (percent) Reasoning low Reported September 22, 2026 Source Evaluation source for GPT-5.6 Luna on OSWorld 2.0 v2026.08.08 Offline — Partial (opens in a new tab) GPT-6 Luna Score 46.7% Evaluator OpenAI Version OSWorld 2.0 v2026.08.08 Offline — Partial Metric OSWorld Partial Score (percent) Reasoning xhigh Reported September 22, 2026 Source Evaluation source for GPT-6 Luna on OSWorld 2.0 v2026.08.08 Offline — Partial (opens in a new tab) Score 31.5% Evaluator OpenAI Version OSWorld 2.0 v2026.08.08 Offline — Partial Metric OSWorld Partial Score (percent) Reasoning medium Reported September 22, 2026 Source Evaluation source for GPT-6 Luna on OSWorld 2.0 v2026.08.08 Offline — Partial (opens in a new tab) Score 52.7% Evaluator OpenAI Version OSWorld 2.0 v2026.08.08 Offline — Partial Metric OSWorld Partial Score (percent) Reasoning max Reported September 22, 2026 Source Evaluation source for GPT-6 Luna on OSWorld 2.0 v2026.08.08 Offline — Partial (opens in a new tab) Score 41.4% Evaluator OpenAI Version OSWorld 2.0 v2026.08.08 Offline — Partial Metric OSWorld Partial Score (percent) Reasoning high Reported September 22, 2026 Source Evaluation source for GPT-6 Luna on OSWorld 2.0 v2026.08.08 Offline — Partial (opens in a new tab) Score 8.3% Evaluator OpenAI Version OSWorld 2.0 v2026.08.08 Offline — Partial Metric OSWorld Partial Score (percent) Reasoning low Reported September 22, 2026 Source Evaluation source for GPT-6 Luna on OSWorld 2.0 v2026.08.08 Offline — Partial (opens in a new tab)