Notes for myself
o‑series reasoning models
- o3: flagship reasoning model with private chain-of-thought.
- o3-mini, o3-mini-high: compact versions with varying reasoning effort levels.
- o3-pro: highly compute‑intensive variant for maximum reliability.
- o4-mini and o4-mini-high: successors to o3-mini, released April 2025; support images and tool‑calling within chain‑of‑thought.
- o3-2025-04-16 is ranked high in LM Arena across various categories.
GPT-4.x
- Multimodal: handles text, images, audio natively.
- Supports tool‑calling and web search within Responses API
- Includes gpt-4o and lightweight gpt-4o-mini variants
- Includes gpt-4.1, gpt-4.1-mini, and ultra‑fast gpt-4.1-nano
- chatgpt-4o-latest-20250326 ranks just below o3-2025-04-16 on LM Arena
Cost Comparison
| Model | Input ($/M tokens) | Output ($/M tokens) | Notes |
|---|---|---|---|
| GPT-4o | ~$2.50 | ~$10.00 | |
| GPT-4o-mini | $0.15 | $0.60 | (Reuters) |
| GPT-4.1 | $2.00 | $8.00 | (Zapier, OpenAI Community) |
| GPT-4.1-mini | $0.40 | $1.60 | (Zapier) |
| GPT-4.1-nano | $0.10 | $0.40 | (Zapier) |
| o3-mini | $1.10 | $4.40 | (Inc.com, API.chat) |
| o3-mini-high | $1.10 | $4.40 | High-effort variant, same pricing tiers (Wikipedia, Inc.com) |
| o3 (standard) | $2.00 | $8.00 | Matches GPT-4.1 pricing now (Zapier, OpenAI Community) |
| o3-pro | $20.00 | $80.00 | Pro reasoning tier (Zapier, OpenAI Community) |
| o4-mini | $1.10 | $4.40 | (OpenAI Platform Compare) |