GPT

Overview

This is a model family overview. For version-specific details, see the individual model entries linked below.

GPT is OpenAI’s core model family for assistant, reasoning, coding, multimodal, and tool-using workloads. The current generation is GPT-5.6: Sol for frontier capability, Terra for balanced production work, and Luna for cost-sensitive high-volume workloads. GPT-5.5, GPT-5.5 Pro, and GPT-5.4 routes remain available for pinned integrations and deliberate migration paths.

This snapshot was rechecked against OpenAI’s public model docs and general-availability announcement on July 10, 2026.

Current Lineup

At this snapshot date, the current GPT picture looks like this:

• Frontier lane: GPT-5.6 Sol is the family flagship for complex coding, professional work, science, design, computer use, and tool-heavy agents. The gpt-5.6 alias routes to Sol.
• Balanced production lane: GPT-5.6 Terra targets strong general performance at half Sol’s standard token price.
• High-volume lane: GPT-5.6 Luna is the lowest-cost GPT-5.6 tier for supporting agents, extraction, classification, and routine transformations.
• Previous current API lanes: GPT-5.5 and GPT-5.5 Pro remain documented, non-deprecated routes for pinned behavior, regression comparisons, and gradual migrations. GPT-5.6 exposes Pro as a reasoning mode on the selected model rather than a separate Pro API slug.
• Still-current API lanes: GPT-5.4 and GPT-5.4 Pro remain important API models for compatibility, cost routing, and teams not yet moving to GPT-5.6. Older GPT-5.2 routes remain compatibility references rather than fresh-build defaults.
• Codex-specialized routes: GPT-5.3-Codex remains API-published but is deprecated in Codex when using ChatGPT sign-in. GPT-5.2-Codex is also deprecated and scheduled for API shutdown on July 23, 2026. Use the current Codex product guidance or GPT-5.6 routes for fresh work.
• Specialized scientific-research lane: GPT-Rosalind is a separate life-sciences research preview for qualified enterprise teams using ChatGPT Enterprise, Codex, and the API.
• Efficiency tiers: GPT-5.4 mini and GPT-5.4 nano are the current recommended starting points for new mini/nano GPT builds, while GPT-5 mini and GPT-5 nano remain older compatibility baselines.
• Compatibility / historical routes: GPT-4o, GPT-4.1, GPT-5, and earlier Codex variants now matter mostly for migration, compatibility, or historical comparison.

Strengths

• Strong general-purpose quality across reasoning, coding, multimodal workflows, and tool use.
• Mature API, playground, eval, and agent tooling ecosystem for production integration.
• Broad tiering for cost-quality routing strategies.
• Large context support in flagship tiers for long-document and agent workflows.
• Frequent model iteration with explicit migration paths between generations.

Tier Guide

• GPT-5.6 Sol: use when the highest capability ceiling matters more than token price.
• GPT-5.6 Terra: use as the broad middle tier for production coding, knowledge work, and agent workflows.
• GPT-5.6 Luna: use for cost-sensitive, high-volume, and supporting workloads.
• GPT-5.5 / GPT-5.5 Pro: still-current API routes for pinned integrations and measured migration, no longer the newest GPT generation.
• GPT-5.4 / GPT-5.4 Pro: still-current premium routes for compatibility, cost routing, or existing production integrations. Both use a 1.05M-token context window in OpenAI’s current API docs, with extra pricing rules for prompts above 272K input tokens.
• GPT-5.2 / GPT-5.2-Pro: older compatibility routes for pinned integrations and regression comparisons.
• GPT-5.3-Codex: still API-published for pinned coding integrations, but deprecated in Codex when using ChatGPT sign-in.
• GPT-5.2-Codex: deprecated compatibility route scheduled for API shutdown on July 23, 2026; OpenAI’s deprecation table names GPT-5.5 as the direct replacement.
• GPT-Rosalind: domain-specific research preview for biology and life-sciences workflows, not a mainstream replacement for general GPT routing.
• GPT-5.4 mini / nano: current efficiency defaults for high-volume, supporting, and subagent work. GPT-5 mini / nano remain useful mainly when an integration is already pinned to the older behavior or price point.
• o-series: separate reasoning-specialized family for targeted deep-analysis routes.

Access

• OpenAI API
• ChatGPT and Codex first-party product surfaces
• Partner platforms exposing OpenAI model endpoints