treyleo16/ gpt-5-6-luna:latest

24 4 days ago

Cost-effective great model by OpenAI

cloud 
ollama run treyleo16/gpt-5-6-luna

Details

4 days ago

347016434c3d · 128kB ·

system
You are ChatGPT, a large language model trained by OpenAI, based on GPT-5.6 Luna. Current date: 2026

Readme

GPT-5.6 Luna

GPT-5.6 Luna is OpenAI’s high-speed, cost-optimized model tier within the GPT-5.6 family (alongside Sol and Terra). Serving as the lightweight nano-tier engine of the suite, Luna is engineered for high-throughput, latency-sensitive applications, high-volume classification, and streaming sub-agent orchestration.

Model Overview

Luna bridges the gap between ultra-low unit economics and frontier-class context processing. It enables developers to run continuous, large-scale agentic pipelines and full-document passes without incurring the cost overhead of flagship reasoning engines.

  • Developer: OpenAI
  • Model Tier: Luna (Fast / High-Throughput Efficiency)
  • Release Date: July 9, 2026
  • Primary Target Use Cases: Real-time interactive assistants, high-volume classification, structured field extraction, sub-agent execution loops, and long-document processing.

Key Capabilities & Performance Profile

Ultra-Low Unit Economics

  • High-Volume Viability: Designed specifically for features that touch every request, user turn, or document pass, keeping operational costs low across high-traffic production APIs.
  • Token Efficiency: Inherits the GPT-5.6 family’s token-reduction optimizations, using fewer generation tokens to reach conclusions on intermediate reasoning steps.

Extended Context & Long Generation

  • 1.1 Million Token Window: Matches the 1,100,000-token context window of its larger siblings (Terra and Sol), allowing full codebases, logs, and lengthy spec sheets to be processed in a single pass.
  • High Output Headroom: Supports output lengths up to 272,000 tokens per request for extended code outputs or document generation.

Sub-Agent Orchestration & Native Tools

  • Agent Pipeline Node: Functions as a low-cost worker model in multi-agent architectures, executing routine classification, tool formatting, and initial task routing before escalating complex edge cases to GPT-5.6 Sol.
  • Responses API Support: Native support for web search, file search, code execution, computer use, and Model Context Protocol (MCP) integrations.

Technical Specifications

Parameter Specification
Model ID gpt-5.6-luna / openai/gpt-5.6-luna
Context Window 1,100,000 tokens (1.1M)
Max Output Tokens 272,000 tokens
Input Modalities Text, Code, Images (Vision)
Output Modalities Text, Code, Structured JSON
Native Features Function Calling, Prompt Caching, Reasoning Tokens, Computer Use, MCP Tools

GPT-5.6 Family Tiering

Tier Role & Capability Target
GPT-5.6 Sol Flagship: Maximum reasoning, top-tier coding, defensive cybersecurity, and deep multi-step architecture.
GPT-5.6 Terra Balanced Workhorse: Mid-tier option balancing intelligence and cost for standard enterprise production traffic.
GPT-5.6 Luna Ultra-Fast Efficiency: High-throughput, low-latency engine designed for cost-sensitive, high-volume workloads.