Up to four models
Hy4 preview vs GPT-6 Luna
On the index (GPT-4 = 100), Hy4 preview scores 173 and GPT-6 Luna 162. Side by side on the same third-party tests, with price and context. The link in your address bar shares this exact comparison.
173154–192
WeightsOpen
ReleasedAug 2026
Price in / out$0.83 / $2.5
Context1.05M
Core tests taken2 of 9
GPT-6 Luna
162146–178
WeightsClosed
ReleasedSep 2026
Price in / out$0.10 / $0.50
Context1.05M
Core tests taken6 of 9
Hy4 preview and GPT-6 Luna are too close to call (11 points apart, within the uncertainty), ahead on 2 of 2 shared core tests, at 6.3× the price.
Test by testthird-party results
Every result as a table
| Test | Hy4 preview | GPT-6 Luna |
|---|---|---|
| Maths | ||
| ProofBench v1.1 | 75.0% | 64.0% |
| Coding | ||
| Terminal-Bench 4.0 | 39.4% | 13.6% |
| Code Migration | 47.4% | 42.5% |
| CyberBench v1.1 | 65.4% | 76.3% |
| IOI | 59.3% | 55.6% |
| ProgramBench | 0.0% | 0.5% |
| Vibe Code Bench v1.1 | 81.0% | 81.7% |
| Agents | ||
| BioMysteryBench | 69.3% | 61.5% |
| SRE Bench | 2.3% | 2.7% |
| Professional work | ||
| Finance Agent (v2) | 55.1% | 49.9% |
| EMB | 58.1% | 68.5% |
| HLAB-Vals-1.1 | 12.5% | 4.2% |
| Legal Research Bench | 45.2% | 30.3% |
| MedCode | 43.3% | 44.7% |
| MedScribe | 83.6% | 83.7% |
| Public Benefits Bench v1.1 | 68.6% | 57.6% |
| Tax Agent Bench | 34.4% | 20.0% |
Tests at least two of these models have taken. Bold rows are the core tests.