# GPT-5
GPT-5 is the fifth-generation [[Large Language Models (LLMs)|LLM]] family from [[OpenAI]]. It succeeds GPT-4 / GPT-4o as OpenAI's flagship closed-weight line and serves as the umbrella for the GPT-5.x minor releases that ship over a generation's lifetime.
## Variants
Minor versions released / referenced as of May 2026:
- **GPT-5** — initial generation release
- **GPT-5.3** — interim refresh
- **GPT-5.3-Codex** — coding-tuned 5.3 variant; long-time default for [[OpenAI Codex]] before 5.5 shipped
- **[[GPT-5.4]]** — frontier flagship through April 2026; the version third parties (e.g., [[Moonshot AI]]'s K2.6 launch) used as their US closed-weight reference
- **GPT-5.5** — current frontier; rolled into [[OpenAI Codex]] surfaces as the recommended default starting late April 2026
- **GPT-5.1-Codex-Mini** — lighter coding-tuned tier; the cost-control answer for routine Codex work that does not need 5.5
Specialty variants (mini / nano / turbo / pro / reasoning-tuned) follow OpenAI's usual tiering; check https://platform.openai.com/docs/models for the current list.
## Where each variant lives
| Surface | Default model |
|---|---|
| ChatGPT (consumer) | GPT-5.5 |
| [[OpenAI Codex]] (CLI / App / IDE / Cloud) | GPT-5.5 (was GPT-5.3-Codex through April 2026) |
| OpenAI API | per-request; pick from the models endpoint |
| Microsoft Azure / Foundry | enterprise tenants typically lag the consumer-side default |
## Positioning
- Closed-weight, hosted via OpenAI API, ChatGPT, and Microsoft Azure / Foundry
- Drives [[OpenAI Codex]] and agentic ChatGPT modes
- Competes with [[Claude Opus 4.7]] ([[Anthropic]]), [[Gemini 3]] ([[Google DeepMind]]), and the open-weight frontier ([[Kimi K2.6]], [[Qwen]])
- The 2026 frontier race is effectively GPT-5.x vs Opus 4.x vs Gemini 3.x vs K2.x
## Why it matters
- The default "current OpenAI model" that every other lab benchmarks against
- ChatGPT's default model under the hood, which means GPT-5.x is what hundreds of millions of users touch daily, regardless of what the enterprise / API crowd prefers
- Minor-version cadence (5.3 → 5.4 → …) is where the interesting quarterly deltas live; the "5 → 6" generational jump will be the larger story when it happens
## Caveats
- OpenAI has not published full model cards at the minor-version level for every variant
- Pricing, context window, and knowledge cutoff vary by surface (ChatGPT vs API vs Azure); check the live docs
- Third-party comparison tables (e.g., in K2.6's launch) are directional, not authoritative
## References
- OpenAI: https://openai.com/
- Model list: https://platform.openai.com/docs/models
## Related
- [[OpenAI]]
- [[GPT-5.4]]
- [[GPT4]]
- [[ChatGPT]]
- [[OpenAI Codex]]
- [[Large Language Models (LLMs)]]
- [[AI Frontier Model]]
- [[AI Foundation Models]]
- [[Claude Opus 4.7]]
- [[Gemini 3]]
- [[Kimi K2.6]]