Hy4 preview scores 46.6 on the Respan Index (#31), ahead of GPT-5.6 Luna at 46.2 (#33). They split the 3 benchmarks both report evenly. GPT-5.6 Luna is 2.8x cheaper per token at list price. GPT-5.6 Luna has the larger context window (1.1M tokens).
| Benchmark | GPT-5.6 Luna | Hy4 preview |
|---|---|---|
| GPQA Diamond | 92.3% | 92.3% |
| Terminal-Bench 4.0 (Vals AI) | 11.6% | 8.1% |
| LMArena Agent (LMArena) | -0.0117 | 0.0395 |
Higher is better. Each score links to where it was published.
* Estimated: no published scores in that category. How the Respan Index works