The OpenAI Model Glossary is a collection of every major model from 📝OpenAI, from GPT-1 in 2018 to the GPT-6 family in September 2026, grouped by family and listed newest first. Most are generations of the 📝GPT architecture; as of September 28, 2026, OpenAI's flagship lineup is GPT-6 Astra, GPT-6 Sol, and GPT-6 Luna, and its models now use named capability tiers rather than version-plus-size names.

Glossary

GPT-6

GPT-6 Astra — gpt-6-astra, released September 3, 2026: OpenAI's most capable model, built for the hardest end-to-end work, with a 1.05M-token context window at $10 / $50 per million tokens.
GPT-6 Sol — gpt-6-sol, released September 22, 2026: the mid-tier GPT-6 model for complex coding and agentic workflows, trained with methods similar to Astra's, at $2 / $10 per million tokens.
GPT-6 Luna — gpt-6-luna, released September 22, 2026: the most efficient GPT-6 model for focused, high-volume tasks, at $0.10 / $0.50 per million tokens.

GPT-5 Generation

📝GPT-5 — released August 7, 2025: a unified system combining a fast model, a deeper "thinking" model, and a real-time router; API snapshot shuts down December 11, 2026.
GPT-5.6 Sol, Terra, and Luna — released July 9, 2026 after a limited June preview: the first OpenAI family to use named capability tiers, from Luna (smallest) through Terra to Sol (flagship).
📝GPT-5.5 — released April 23, 2026 in ChatGPT and Codex, with API access the next day.
GPT-5.4 — released March 5, 2026: the first mainline OpenAI model with native computer use, followed by GPT-5.4 mini and nano on March 17.
GPT-5.2 — released December 11, 2025: the GPT-5 flagship aimed at professional knowledge work.
GPT-5.1 — released November 12, 2025: added adaptive reasoning and a warmer default tone.

Codex Models

GPT-5.3-Codex — released February 5, 2026: the coding-specialized model between GPT-5.2 and GPT-5.4; API shut down July 23, 2026.
GPT-5-Codex — released September 15, 2025: a version of GPT-5 optimized for agentic coding in 📝Codex.
codex-1 — released May 16, 2025: the model powering the Codex cloud agent's research preview in ChatGPT.
Codex (original) — released August 10, 2021: a GPT-3 descendant fine-tuned on code that powered the first 📝GitHub Copilot; retired March 2023.

Reasoning (o-series)

o3-pro — released June 10, 2025: a higher-compute version of o3; API shutdown December 11, 2026.
o3 and o4-mini — released April 16, 2025: the first reasoning models able to use every tool in ChatGPT agentically; o3 shuts down December 11, 2026.
o3-mini — released January 31, 2025: the first reasoning model offered to free ChatGPT users.
o1 — released December 5, 2024: the first full reasoning model, trained to think before responding; API shutdown October 23, 2026.
o1-preview and o1-mini — released September 12, 2024: OpenAI's first reasoning models, retired in 2025.

GPT-4 Generation

📝GPT-4-series — OpenAI's fourth GPT generation, from GPT-4 (2023) to GPT-4.1 (2025), retired from ChatGPT on February 13, 2026.
GPT-4.5 — released February 27, 2025: a large research-preview model, retired from the API July 14, 2025.
GPT-4.1, GPT-4.1 mini, and GPT-4.1 nano — released April 14, 2025: API-only models that introduced a 1M-token context window.
GPT-4o mini — released July 18, 2024: the cost-efficient version of GPT-4o.
GPT-4o — released May 13, 2024: the "omni" model handling text, audio, and image input and output in one network.
GPT-4 Turbo — released November 6, 2023: extended GPT-4 to a 128K-token context; API shutdown October 23, 2026.
GPT-4 — released March 14, 2023: OpenAI's first multimodal flagship, accepting text and image input.

Early GPT

GPT-3.5 and ChatGPT — released November 30, 2022: the model behind 📝ChatGPT's research preview, which set off mainstream adoption of generative AI.
📝GPT-3 — released June 11, 2020: the first GPT model OpenAI commercialized broadly, through a private API beta.
GPT-2 — released February 14, 2019: staged release, with the full 1.5-billion-parameter model following in November 2019.
GPT-1 — released June 11, 2018: the first generative pre-trained transformer, introducing unsupervised pre-training plus supervised fine-tuning.

Open-Weight

gpt-oss-120b and gpt-oss-20b — released August 5, 2025: OpenAI's first open-weight language models since GPT-2, published under the Apache 2.0 license.

Specialized and Trusted Access

GPT-5.6-Cyber — released August 2026: a cybersecurity model based on GPT-5.6 Sol, available only through the Daybreak Red trusted-access tier.
GPT-5.5-Cyber — 2026: the earlier Daybreak Red cyber model; gpt-5.4-cyber shuts down October 1, 2026.
GPT-Rosalind — research preview from April 2026: a life-sciences reasoning model for drug discovery and biology, available to qualified organizations through trusted access.

Image

GPT Image 2.5 Sunburst and Flare — released September 8, 2026: Sunburst for quality-critical editing and Flare as the fast default, both successors to GPT Image 2.
GPT Image 2 — released April 21, 2026: OpenAI's first image model with thinking capabilities.
GPT Image 1 — released April 2025 in the API after GPT-4o's native image generation debuted in 📝ChatGPT Images on March 25, 2025; API shutdown October 23, 2026.
DALL·E 3 — released October 19, 2023: integrated into ChatGPT; API retired May 12, 2026.
DALL·E 2 — released April 6, 2022: higher-resolution, more realistic image generation; API retired May 12, 2026.
DALL·E — released January 5, 2021: OpenAI's first text-to-image model.

Video

Sora 2 — released September 30, 2025: OpenAI's flagship video-and-audio generation model; discontinued, with the API shut down September 24, 2026.
📝Sora — previewed February 15, 2024 and launched December 9, 2024: OpenAI's first text-to-video model.

Voice and Audio

GPT-Live-1 and GPT-Live-1 mini — released July 8, 2026 in ChatGPT and September 10, 2026 in the API: full-duplex voice models that listen and speak at the same time.
GPT Transcribe and GPT Live Transcribe — released July 28, 2026: successors to the Whisper and GPT-4o transcription lines.
📝Whisper (OpenAI) — released September 21, 2022: OpenAI's open-source speech-recognition model; whisper-1 shuts down February 26, 2027.