OpenAI splits GPT-5.6 as labs compete on price

OpenAI released GPT-5.6 as three models rather than a single flagship system: Sol for frontier reasoning and long-horizon agent work, Terra for everyday tasks at roughly GPT-5.5-level quality at about half the cost, and Luna as the fastest and cheapest option. The shift frames model choice around cost and workload fit, while adding some complexity for users deciding which tier to use.

OpenAI also introduced GPT-Live-1 and a mini version, full-duplex voice models designed to listen and speak at the same time. The release points toward assistants that can handle interruptions and adjust mid-conversation rather than relying on turn-based exchanges.

Other launches reinforced the same market shift. xAI released Grok 4.5, Google made Gemini 3.5 Flash the global default in AI Mode, and Base44 shipped Base 1, an in-house model trained on tens of millions of real app-building sessions. CNBC reported that US companies are increasingly considering Chinese models such as DeepSeek and Z.ai as OpenAI and Anthropic costs rise, underscoring a broader move from capability races to affordability and distribution.