OpenAI launched the GPT-5.6 model family featuring Sol, Terra, and Luna. Sol outperforms rivals in coding benchmarks at under half the API cost. Terra matches predecessor performance at half the price. Luna is 80% cheaper than Sol. Technical stack optimizations including speculative decoding and cache improvements boost token efficiency by over 15%. Analysts say this shift toward maximum efficiency will accelerate AI deployment in complex scenarios.