Skip to main content

GPT-6 Luna

gpt
Run #430 · Formula v7 · Judge v6.5 · Benchmark v7

Best value

79.6
Overall Score
#1 / 15
Current Rank
11-16 05:16 SGT
Last Evaluated
Recommended Core Overall 76.03
Normal Updated 11-18 03:31

Core Dimensions (v6) v6

Code Execution 78.5 Grounding 73 Engineering Judgment 86.5 Task Communication 79.4 Integrity Rating 78.3
PASS
Integrity
Integrity Score 78.30
Code Execution
78.5
Grounding
73
Engineering Judgment
86.5
Task Communication
79.4
Integrity Rating
78.3
Show v5 legacy dimensions

Legacy Dimensions (v5) legacy

Code Execution 78.1 Knowledge 82.3 Long Context 73 Value 100 Stability 36.1 Availability 99
Code Execution
78.1
Knowledge
82.3
Long Context
73.0
Operational Metrics
Value
100.0
Stability
36.1
Availability
99.0

WDCD Compliance Test Pilot

78.40
WDCD Score
#9
Compliance Rank / 15
Three-Round Performance
R1 Acknowledgment
1.00/1
R2 Resistance
0.50/1
R3 Integrity
1.00/2

View full WDCD compliance rankings

Frequently Asked Questions

How does GPT-6 Luna perform on the YZ Index benchmark?

In the latest public YZ Index evaluation on 2026-11-16, GPT-6 Luna scored 76 overall (out of 100), ranking #6 among 15 models. The score aggregates four core dimensions — real code sandbox execution, grounding, engineering judgment, and task communication — with 100% rule-based scoring.

How good is GPT-6 Luna at coding?

GPT-6 Luna scores 78.5 on the Execution dimension, ranking #8 of 15 models. This dimension actually runs model-generated programs in an isolated sandbox to verify compilation, runtime correctness, and edge-case handling — no model-as-judge scoring.

Can GPT-6 Luna keep instruction constraints over long conversations?

In the WDCD instruction-decay test (2026-11-18), which measures instruction compliance under multi-turn pressure, GPT-6 Luna scored 78.4 (out of 100), ranking #9 of 15 tested models. WDCD applies progressive distraction and social-engineering pressure, with 100% rule-based scoring.

How much does the GPT-6 Luna API cost?

GPT-6 Luna's API is priced at $0.1 per million input tokens and $0.5 per million output tokens (official pricing pages are verified regularly). See the Value dimension on this page for price-performance.

How often is this benchmark data updated?

The YZ Index runs a full evaluation weekly and sampled evaluations daily; this page updates automatically with every public run. Current data comes from Run #430 on 2026-11-16; see the trend chart for history.

Recent Changes

communication_raw -5.2 GPT-6 Luna:任务表达 -5.2

Score Trend

0 20 40 60 80 100 10-05 10-12 10-19 10-26 11-02 11-09 11-16 vv6.5

v6 scores are from the latest evaluation run

Back to Model List