OpenAI unveils GPT‑5.6 three‑tier pricing as DeepSeek Flash claims 200× lower cost

OpenAI launched the GPT‑5.6 family on July 9, 2026 with a three‑tier pricing menu that lets developers route tasks to the model best suited for the job. The lineup includes Sol (flagship), Terra (balanced) and Luna (fastest, cheapest). Pricing per million tokens is $5 input / $30 output for Sol, $2.5 input / $15 output for Terra, and $1 input / $6 output for Luna, with additional cache‑write charges. Access to the models varies by subscription plan; Sol is available to higher‑tier users, while Terra and Luna are offered to lower‑tier plans, and standard ChatGPT conversations use Sol by default. OpenAI kept overall ChatGPT subscription prices unchanged.

In a separate analysis, DeepSeek Flash is positioned as a cost‑effective alternative to OpenAI’s GPT‑4. Its base price is about $0.14 per million tokens, and a high cache‑hit rate (≈90%) reduces the effective cost to roughly $0.014 per million tokens. This represents a 200‑fold saving compared with GPT‑4’s $30 input / $60 output rates. DeepSeek Flash is reported to handle about 90% of typical tasks—customer support, content generation, code assistance—with comparable quality, reserving GPT‑4 for the most complex reasoning and creative writing needs.