| Model profile | ||
|---|---|---|
| Model | Size | Approximate size |
| GPT 5.6 LunaModel source: provider1,050,000 tokens max native context | not available | not available4-bit quantization |
Ranks frontier models based on real use across front-end design, one-shot capabilities, cost, speed, and subscriptions, placing Fable 5 at the top.
Tests GPT-5.6 Luna performance on browser OS, C++ games, FPS dev, multimodal coding, 3D CAD models, watch frontends, Chrono City C++, HQ Street Yeet, and result improvement.
Benchmarks GPT-5.6 Sol, Terra, and Luna against Claude Fable 5, Grok 4.5, and GLM 5.2 via Artificial Analysis, OSWorld, Terminal-Bench, coding, frontend generation, and cost evaluations.