GPT-5.6 Terra

GPT-5.6 Terra is OpenAI’s balanced, mid-tier workhorse model within the GPT-5.6 family. Positioned directly between the high-capacity Sol tier and the ultra-fast Luna tier, Terra delivers high reasoning, coding, and agentic capabilities at roughly half the operating cost of Sol.

Model Overview

Terra is engineered to serve as the default production model for general enterprise automation, daily coding tasks, and multi-turn agent workflows. It provides a strong performance-to-cost ratio, allowing teams to deploy complex agent loops at scale without incurring flagship tier overhead.

• Developer: OpenAI
• Model Tier: Terra (Balanced Workhorse)
• Release Date: July 9, 2026
• Primary Target Use Cases: Production web applications, routine codebase maintenance, automated customer operations, multi-turn agentic task loops, and structured document processing.

Key Capabilities & Performance Profile

Balanced Agentic Coding & Reasoning

• Cost-Efficient Performance: Achieves high benchmark accuracy across coding and terminal tasks while consuming substantially fewer generation tokens than previous-generation models.
• Agentic Execution: Excels at file modification, API tool coordination, and multi-step terminal workflows in persistent coding environments like Codex.

Extended Context & Prompt Caching

• 1 Million Token Window: Supports up to 1,050,000 tokens of input context natively, handling large source repositories, full documentation sets, and deep chat logs in a single request.
• Explicit Cache Breakpoints: Integrates with OpenAI’s prompt caching infrastructure. Repeated system prompts and workspace context receive significant cost reductions on cache hits.

Technical Specifications

GPT-5.6 Family Positioning