GPT-5.6 Luna
Routed in Coreby OpenAI
Efficient everyday agent work and quick iteration.
Intelligence51Artificial Analysis Index
Average task cost$0.21Observed benchmark task
Cost per million$1 in$6 out
Output speed188tokens per second
Evidence
Open source What we know
- Availability
- Available through Core's model router
- Core fit
- 50 / 100 · Efficient
- Public intelligence
- 51 · max effort
- Benchmark snapshot
- Jul 2026
- Input price
- $1 per 1M tokens
- Output price
- $6 per 1M tokens
- Cache read
- $0.1 per 1M tokens