Public
benredmond / apex
Benchmark updated: 8/13/202630 Tasks
Custom repository benchmark.
Languages
HTML72.5%TypeScript19.3%JavaScript7.9%Shell0.2%CSS0.1%Dockerfile<0.1%
Harness | Input / Output Cost | ||||||
|---|---|---|---|---|---|---|---|
1 |  Mini-SWE-agent | 19 / 30 | $0.09 | $0.2/$1.2 | 6m04s | ||
2 | Mini-SWE-agent | 18 / 30 | $0.45 | $2/$12 | 4m10s |
Key Takeaways
- GPT-5.6 Luna with Mini-SWE-agent scores 63.33% versus 60% for GPT-5.6 Terra with Mini-SWE-agent.
- GPT-5.6 Terra with Mini-SWE-agent is faster at 249.87 seconds versus 363.72 seconds, while GPT-5.6 Luna with Mini-SWE-agent costs less per test: $0.09 versus $0.45.
Cost Analysis
Cost / Test vs. Accuracy
ACCURACYCOST
Average Token Use / Test
Token Usage
InputOutputReasoningCache readCache write
GPT-5.6 Luna
GPT-5.6 Terra
Cost is the clearest tradeoff in this comparison. GPT-5.6 Luna leads at 63.33% for $0.09 per test. No other model in this comparison is cheaper.
Latency Analysis
Latency vs. Accuracy
ACCURACYLATENCY
Average Response Time / Test
Response Time
GPT-5.6 Luna
GPT-5.6 Terra
Latency separates several models with similarly strong scores. GPT-5.6 Luna leads at 63.33%, while GPT-5.6 Terra is fastest at 4m 10s with 60.00% accuracy.