GPT 5.6 Terra

OpenAI announced a limited preview of the GPT-5.6 series on June 26, 2026, explicitly naming Terra as a balanced tier within the family: "We're beginning a limited preview of the GPT-5.6 series: Sol, our flagship model; Terra, a balanced model for everyday work; and Luna, a fast and affordable model." The announcement The launch post also outlines the rollout path: Terra (alongside Sol and Luna) is planned for general availability "in the coming weeks," and OpenAI is working with the U.S. government on a cyber Executive Order framework during the preview period. This is the strongest first-party evidence confirming Terra as a named

A dedicated benchmark dossier for GPT-5.6 Terra reports pricing of $2.50 input / $15.00 output per million tokens and provides detailed evaluation results. Key scores include ARC-AGI-1 at 96.5% (rank 10/97), ARC-AGI-2 at 83.9% (rank 11/99), and Tau3-Banking at 40.2% (rank 15/17). RuneBench metrics show Attack XP/min at The page enables head-to-head comparison with sibling models including GPT-5.6 Sol, Claude Fable 5.1, Claude Opus 5, and Kimi K3 across shared benchmarks, giving developers a concrete view of Terra's competitive standing within the GPT-5.6 family and against other frontier models. Evaluation dates range from late Augus

An independent benchmark review dated July 14, 2026 evaluates GPT-5.6 Terra specifically. Terra scores 77.4 on the Artificial Analysis Coding Agent Index, 63.4% on SWE-Bench Pro (1.2 points below Sol at 64.6%), and 87.4% on Terminal-Bench 2.1. The review concludes "Terra scores within 3 points of Sol on most benchmarks The review characterizes Terra as suited for "development loops" and scoped implementation work, contrasting it with Sol for "production gates" and Luna for "volume." Terra's pricing places it as the mid-tier between Sol ($5/$30) and Luna ($1/$6), and it supports the same reasoning modes (none, low, medium, high, xhigh

A Medium article by The Ai Consultancy recounts that on June 26, 2026 OpenAI announced GPT-5.6 Sol, Terra, and Luna in a limited preview accessible only to approved partners via the API, Codex, or both, with ChatGPT excluded and no public waitlist or application route. Thirteen days later, on July 9, OpenAI declared th The same article highlights that GA introduced a broader set of evaluations, folded GPT-5.6 into a larger ChatGPT Work launch, and incorporated the Codex desktop app into a new ChatGPT desktop application. It notes that several key facts such as pricing tiers and reasoning settings remained unchanged across stages, and

A third-party analysis published July 9, 2026 (reviewed July 28, 2026) characterizes the GPT 5.6 family as a three-tier release from OpenAI: GPT 5.6 Sol for the hardest work, GPT 5.6 Terra as the balanced production default, and GPT 5.6 Luna for cost-controlled high-volume tasks. It positions Terra as the sensible prod The article frames the primary business value of the release as the family design rather than Sol alone: Sol for peak capability, Terra for everyday agents, Luna for cost-sensitive scale. It draws on OpenAI's GPT 5.6 launch post, the GPT 5.6 System Card, OpenAI model docs, and pricing pages, and explicitly notes that b

The llm-stats.com model page for GPT-5.6 Terra (published July 9, 2026) ranks the model 16th on the site's composite LLM Stats Score, with capability tiers placing it in the Top 10% for Coding (8 of 267) and Search (8 of 90), average in categories like 3D, Chat, Long Context, and Tool Calling, and below the top half in The page records a Quality Tracker reading of +1.29σ with 88 votes over the prior 7 days, marked Stable, and provides per-dataset deltas including Chat +2.74σ (42 votes), Animations +2.61σ (6 votes), and Audio +1.70σ (3 votes). It also lists individual benchmark scores such as a perfect 1.00/1 pass rate on the Connecto

CodeRabbit published a July 9, 2026 engineering review positioning GPT-5.6 Sol and Terra for coding agents and code review, noting that Sol follows through better on long-horizon coding tasks including messy repo work, long checklists, and follow-up fixes, while Terra serves as the cheaper lane for scoped implementatio CodeRabbit frames GPT-5.6 as a family with capability tiers rather than a single model, recommending Sol for long-horizon coding, Terra for scoped work, and Luna as the fastest, lowest-cost tier for high-volume tasks where speed and cost dominate. The review also emphasizes that teams should measure cost per solved tas

VentureBeat's June 26, 2026 coverage explicitly names GPT-5.6 Terra and frames it as the high-volume business tier for tasks like customer support, internal tools, and document analysis, distinct from Sol's complex coding/security focus and Luna's fast/cheap positioning. The article reports that Terra and Sol set new b The piece adds release-process context absent from the launch post, noting OpenAI previewed plans and capabilities with the government and is staging access through trusted partners before broader public availability. It situates the GPT-5.6 rollout alongside an export-control order against Anthropic and the removal of

A third-party timeline on hidekazu-konishi.com documents the OpenAI GPT model lineage from GPT-1 in 2018 through the GPT-5 generation in 2025–2026, placing the GPT-5.6 release within the broader context of multimodal omni models, the o-series reasoning line, and the unified GPT-5 generation. The page tracks capability The timeline entry records the deprecation and retirement lifecycle for OpenAI models as listed on OpenAI's official model deprecations page, and situates GPT-5.6 as the latest step in the GPT lineage alongside companion coverage of Anthropic Claude, AWS Generative AI history, and the Amazon Bedrock model catalog. It w