GPT-5.6 Luna

OpenAI announced a limited preview of the GPT-5.6 series on June 26, 2026, explicitly introducing Luna as "a fast and affordable model" alongside the flagship Sol and the balanced Terra tier. According to the post, Luna is positioned to bring strong capability at OpenAI's lowest cost within the family, with general ava The preview post frames GPT-5.6 Sol as the strongest model yet and notes that GPT-5.6 Luna is part of the same family rollout with strengthened safety protections for higher-risk activity, sensitive cyber requests, and repeated misuse. OpenAI indicated that an expanded suite of evaluation results and additional safety

DataCamp's comparison article corroborates that GPT-5.6 Luna is the cheapest tier in OpenAI's GPT-5.6 family, which reached general availability on July 9, 2026, alongside flagship Sol and mid-tier Terra. Luna offers a 1,050,000-token context window with a 128,000-token output ceiling and inherits Programmatic Tool Cal The piece further documents that on July 30, 2026, OpenAI cut Terra's price by 20% to $2.00 input / $12 output per 1M tokens, framing Luna as the high-volume cost-efficient choice and Terra as the balanced mid-tier. This third-party coverage provides useful comparative context for developers selecting between databrick

OpenAI's July 30, 2026 product post announces an 80% price reduction for GPT-5.6 Luna and a 20% reduction for GPT-5.6 Terra, effective immediately across OpenAI usage and reflected in paid Codex and ChatGPT Work subscription billing. The post positions Luna as "our fastest and most affordable model," capable of using t The same announcement introduces Fast mode in the API for GPT-5.6 Sol, delivering up to 2.5x faster speeds than Standard processing at twice the price, with backward compatibility for existing priority-tagged requests. These pricing and capability changes form the upstream context for the Databricks-hosted databricks-g

Databricks' official July 2026 release notes (last updated July 14, 2026) confirm that OpenAI GPT-5.6 Sol, GPT-5.6 Terra, and GPT-5.6 Luna became available as Databricks-hosted models on Model Serving via Foundation Model APIs pay-per-token on July 9, 2026. The same notes also cover adjacent July platform updates, incl The release-note entry directly establishes that GPT-5.6 Luna (modelKey databricks-gpt-5-6-luna, family gpt-luna) is now reachable as a first-party Databricks-hosted endpoint, eliminating the need for customers to call OpenAI directly and bringing GPT-5.6 into the same governance and billing surface as other Foundation

The Databricks Foundation Model APIs limits and quotas page (last updated July 9, 2026) explains the rate-limit regime that governs pay-per-token endpoints such as `databricks-gpt-5-6-luna`, including input tokens per minute (ITPM), output tokens per minute (OTPM), and queries per hour (QPH), with the most restrictive For teams integrating GPT-5.6 Luna into production pipelines, this limits page is the operational reference for sizing concurrency and throughput budgets under pay-per-token billing, and for deciding when to graduate a workload to provisioned throughput to avoid 429 throttling. Pre-admission validation means clients sh

The Databricks Foundation Model APIs supported-models page (last updated July 9, 2026) is the canonical reference for GPT-5.6 Luna on Databricks, listing the endpoint name `databricks-gpt-5-6-luna` with text and image inputs, a 1.05M total token context window, and a 128K maximum output token limit. The page describes For practitioners, the supported-models entry turns the earlier release-note announcement into actionable API details: the exact endpoint string to use in requests, the multimodal input surface for vision-enabled RAG and agent workflows, and the very large 1.05M-token context that materially expands retrieval and long-

Third-party benchmark aggregator llm-stats.com maintains a dedicated tracking page for GPT-5.6 Luna, listing it at rank 32 on the composite LLM Stats Score with a score of 45.1 and a blended price around $0.25 per million tokens as of mid-September 2026. The page places Luna in the Top 10% for Coding (21 of 267 tracked The llm-stats page reports a perfect 1.00/1 score on OpenAI's Connectors production benchmark for GPT-5.6 Luna, alongside tracked performance on HealthBench Consensus and other evaluations, with a Quality Tracker showing stable community voting within roughly ±2σ over the preceding 30 days. Conversation-depth data indi

OpenAI's July 9, 2026 general availability post for the GPT-5.6 family explicitly names GPT-5.6 Luna as the most cost-efficient tier and reports that, together with Terra, it outperforms Claude Fable 5 at around one-sixteenth the cost. The post introduces a new highest-capability "ultra" setting that coordinates multip The post includes two subsequent updates affecting Luna: on July 30, 2026, OpenAI reduced the price of GPT-5.6 Luna by 80% (and GPT-5.6 Terra by 20%), and on August 21, 2026, OpenAI dropped GPT-5.6 Sol API and credit pricing by over 20% for three months. OpenAI framed the family as delivering stronger performance per d