OpenAI Chat Model Cheat Sheet - Decoding the Chaos
10 October 2025 by Brady Stroud
OpenAI’s model names (e.g., o4, 4o, 4.1) are confusing and change often.
This is my quick reference for picking the right model for any task 🧠⚡️
This is a living document - if you spot mistakes or missing models,
please make a PR!
This page lives in GitHub Markdown via TinaCMS 🦙
OpenAI Chat Model Cheat-Sheet (Updated October 2025)
Legend:
- 🟢 current & recommended
- 🟡 active but phasing out
- 🔴 legacy / deprecated
| Status | Model | Release / Change | Strengths / Best Use-Cases | Trade-offs / Notes |
| 🟢 | GPT-5 | Aug 2025 | Unified “smart” (fast) + “thinking” (deep) routing model. Best for general work, reasoning, coding, and RAG. | Variable latency; cost scales with reasoning depth. |
| 🟢 | GPT-4.1 | 14 Apr 2025 | Flagship API model - top coding, logic, and 1 M token context. | Pricier than mini/nano. |
| 🟢 | GPT-4.1 mini | 14 Apr 2025 | ~90 % of 4.1 accuracy at lower price/latency; great default. | Slightly less reasoning depth. |
| 🟢 | GPT-4.1 nano | 14 Apr 2025 | Ultra-fast and cheap; still supports 1 M context. Perfect for classification / autocomplete. | Loses nuance and creativity. |
| 🟡 | GPT-4o / 4o mini | May-Jul 2024 | Multimodal (text + image + audio) real-time chat. | Slightly weaker reasoning than 4.1 series. |
| 🟡 | GPT-4 Turbo / 3.5 Turbo | 2023-2024 | Older “Turbo” line optimized for speed & cost. | Sunsetting by late 2025. |
| 🟡 / 🔴 | GPT-4.5 (“Orion”) | Feb 2025 → retired mid-2025 | Bridge between 4o and 4.1; good reasoning. | Deprecated for GPT-5 and 4.1. |
| 🔴 | GPT-4 (base) | 14 Mar 2023 | Original 4-series model (text-only). | Retired Apr 2025. |
| 🔴 | GPT-3.5 (base) | 2022 | Early RLHF (ChatGPT v1). | Obsolete; small context. |
| 🔴 | GPT-3 / 2 / 1 | 2018 - 2020 | Historical models for context. | Too limited for modern tasks. |
Decoding the Names
- GPT = Generative Pre-trained Transformer
- Major number (3, 4, 5) → architecture generation
- .1 / .5 → mid-cycle refresh or fine-tune
- o → omni (multimodal I/O line)
- mini / nano / high → size vs accuracy trade-off
- Turbo → throughput-optimized snapshot
- MMDD in ID (e.g.
gpt-4-0613) → training-snapshot date