OpenAI: GPT-5.6 arrives in three variants — Sol, Terra and Luna, with multi-agent orchestration and same-day availability in GitHub Copilot
GPT-5.6 is OpenAI's new model family with three variants: Sol (flagship for complex reasoning), Terra (balanced), and Luna (high-volume, cost-efficient). It introduces Programmatic Tool Calling, explicit prompt cache control, persisted reasoning and multi-agent orchestration in beta — available from day one in GitHub Copilot.
This article was generated using artificial intelligence from primary sources.
On July 9, 2026, OpenAI released GPT-5.6, a new generation of its model family that for the first time arrives in three clearly separated variants. GPT-5.6 Sol is the flagship for the most demanding reasoning, GPT-5.6 Terra balances intelligence and cost, and GPT-5.6 Luna targets high-volume tasks where cost per token is decisive. The alias gpt-5.6 in the API routes to Sol.
What do the three variants change for development teams?
Instead of a single model for everything, OpenAI now offers an explicit choice by cost-capability ratio — an approach that Anthropic (Opus/Sonnet/Haiku) and Google (Pro/Flash) standardized earlier. For development teams this means GPT-5.6 Luna can handle classification and extraction at high volume, while Sol takes on agentic tasks over large code repositories. OpenAI describes the release as “more intelligence from every token and stronger performance per dollar” compared to GPT-5.5.
New API capabilities
The official API changelog lists four key additions. Programmatic Tool Calling gives programmatic control over tool invocation. Explicit prompt cache control lets development teams decide what gets cached rather than relying on automatic behavior. Persisted reasoning with configurable effort levels preserves the reasoning chain between calls. Multi-agent orchestration in the Responses API (beta) is OpenAI’s first native support for coordinating multiple agents — functionality that development teams previously built themselves through frameworks like LangChain. Models also now accept images in their native dimensions (original detail).
Same-day Copilot rollout
GitHub announced the same day that all three GPT-5.6 variants are available in GitHub Copilot — in VS Code, JetBrains IDEs, and mobile apps, with usage-based billing. This is the broadest simultaneous rollout of a new GPT generation in developer tools to date: GPT-5.5 waited weeks for the same availability. Sol is positioned in Copilot for complex reasoning over large codebases, Terra as the default model, and Luna for lighter tasks.
Race context
The announcement comes in a week when Meta launched Muse Spark 1.1 with a million-token context, and xAI released Grok 4.5 at $2 per million input tokens. The frontier race in mid-2026 is visibly shifting from individual models to complete families with built-in agentic orchestration.
Frequently Asked Questions
- What are the differences between GPT-5.6 Sol, Terra and Luna?
- Sol is the flagship for the most complex reasoning over large amounts of context (alias gpt-5.6), Terra offers a balanced intelligence-to-cost ratio for everyday work, and Luna is optimized for high-volume, cost-sensitive tasks.
- What is Programmatic Tool Calling in GPT-5.6?
- Programmatic Tool Calling allows the model to invoke tools programmatically, giving development teams deterministic control over when and how tools are executed, rather than relying solely on the model's own choices.
- Where is GPT-5.6 available from day one?
- Through the OpenAI API (Responses API with multi-agent beta orchestration) and in GitHub Copilot for VS Code, JetBrains, and mobile apps, with usage-based billing.
Sources
Related news
Google: SensorFM — foundation model trained on one trillion minutes of wearable data wins on 34 of 35 health tasks
Meta: Muse Spark 1.1 brings multimodal reasoning with one million token context and sub-agent coordination via Model API
Microsoft: Aurora 1.5 open-source model outperforms ECMWF ensemble on 88.9% of variables — new standard in AI weather forecasting