Tabbit
OpenAI now lists GPT-6 Luna with API model ID `gpt-6-luna`. Its published specs include a 1.05M-token context window, up to 128K output tokens, and reasoning efforts of none, low, medium (default), high, xhigh, and max. Standard pricing is $0.10 per million input tokens, $0.01 for cached input, and $0.50 per million output tokens. Availability in Tabbit depends on your account’s live model picker.
Use in Tabbit GPT-6 Luna
GPT-6 Luna · Model overview
A compact view of reviewed task guides, public evaluations, and evidence boundaries. Client access still depends on your current account.
Model access and permissions must be checked in the current Tabbit account.
Extract structured data, build a visual prototype, or start a coding task.
OpenAI provides ready-to-use prompt excerpts for initiative, instruction priority, writing style, delegation, and testing, and recommends evaluating them with the selected GPT-6 model and workload.
The model page confirms the API model ID gpt-6-luna and lists its reasoning efforts, endpoints, input modalities, and Responses API tool support.
The guide presents reusable prompt engineering methods for message roles, the instructions parameter, context, examples, and formatting, and recommends using evaluation sets to validate prompt behavior after model upgrades.
OpenAI outlines a process for monitoring cache hits, diagnosing cache misses, keeping prompt prefixes and tool definitions stable, adjusting reasoning effort, and prewarming the cache.
Public results use different versions, tiers, and harnesses; unknown values stay unknown.
OpenAI positions GPT-6 Luna as a low-cost, high-throughput model and reports improvements over its predecessor on AutomationBench, DeepSWE, and OSWorld, but the public page does not provide complete per-task data or all the configurations needed for replicatio.
Artificial Analysis's own evaluations show GPT-6 Luna max delivering results broadly similar to its predecessor at much lower task cost, while scoring slightly lower on the Coding Agent Index and showing knowledge-work deliverable quality issues on GDPval and .
Artificial Analysis's summary of six effort levels shows substantial differences in Luna's performance, speed, and cost per task: max scores highest, while low costs the least.
OpenAI
Review GPT-6 Luna’s official specifications, reusable prompts, and evaluation evidence to see where it fits high-volume tasks and verify each conclusion at its source.