On July 9, 2026, OpenAI launched its new GPT-5.6 model family, made publicly available after a limited-access phase that started June 26 for select partners. The standout feature is the three-tier structure OpenAI calls Sol, Terra and Luna.

🧠 Three models for three needs

GPT-5.6 isn’t a single model but a family designed for different use cases:

  • Sol - the most powerful model, aimed at complex reasoning, coding and agentic workflows
  • Terra - a balance between capability and cost for everyday work
  • Luna - maximum speed for lightweight, cost-sensitive tasks

Sam Altman described Sol as the “best coding model” OpenAI has ever shipped, with 54% better token efficiency on agentic coding tasks compared to previous models.

🤖 ChatGPT Work: the agent that completes whole jobs

Alongside GPT-5.6, OpenAI introduced ChatGPT Work, a service that uses Sol as its core engine. Unlike traditional conversational assistants, ChatGPT Work is designed to autonomously complete entire jobs: researching prospects, drafting emails, scheduling them and logging activity in CRMs, all without constant human intervention.

💰 API pricing

Model Input (per million tokens) Output (per million tokens)
Sol $5.00 $30.00
Terra $2.50 $15.00
Luna $1.00 $6.00

Sol is significantly pricier than its predecessors, reflecting its position as a frontier model for heavy professional use.

📊 Benchmark performance

According to OpenAI, Sol outperforms previous models and competitors in coding, knowledge work, cybersecurity and science, using fewer tokens and a lower estimated cost per completed task. Independent benchmarks place Sol among the most capable models on the market as of July 2026.

👥 What this means for ChatGPT users

Plus and Team users get Terra as their default model. Sol is available on higher tiers and via the API. Luna is positioned as the option for developers who need high speed at low cost.

GPT-5.6 marks OpenAI’s biggest leap of 2026 and cements its position against competitors like Anthropic, Google and xAI.