Altman says GPT‑5.6 boosts token efficiency by 54% for agentic coding

Sam Altman, CEO of OpenAI, commented on the upcoming GPT‑5.6 model. He noted that GPT‑5.6 achieves a 54% improvement in token efficiency. The efficiency gain applies specifically to agentic coding tasks. Token efficiency measures the amount of text processed per computation. Higher efficiency can reduce compute costs for developers. The claim was made in a public statement on July 9, 2026. OpenAI has not released detailed benchmarks for the model yet. Industry observers will watch how the improvement impacts AI development tools.