OpenAI社の比較、推奨情報が乗っているので、まとめてみました。
Which model should I use?
We generally recommend that you use either gpt-4 or gpt-3.5-turbo. Which of these you should use depends on the complexity of the tasks you are using the models for. gpt-4 generally performs better on a wide range of evaluations. In particular, gpt-4 is more capable at carefully following complex instructions. By contrast gpt-3.5-turbo is more likely to follow just one part of a complex multi-part instruction. gpt-4 is less likely than gpt-3.5-turbo to make up information, a behavior known as "hallucination". gpt-4 also has a larger context window with a maximum size of 8,192 tokens compared to 4,096 tokens for gpt-3.5-turbo. However, gpt-3.5-turbo returns outputs with lower latency and costs much less per token.
We recommend experimenting in the playground to investigate which models provide the best price performance trade-off for your usage. A common design pattern is to use several distinct query types which are each dispatched to the model appropriate to handle them.
GPT-4
- より複雑なタスクをこなせる
- ハルシネーションが起こる可能性がより低い
- テキストだけではなく画像も処理できる
GPT3.5
- 応答速度が速い
- 安い（GPT-4の1/10）
価格の比較（1ドル150円計算）
- 1000トークンあたりの費用
- 1トークンは日本語だと1文字、英語だと１単語
-
1106やpreviewのほうが安い
頭の良さの比較
https://arxiv.org/abs/2303.08774
- どの科目でもGPT-4のほうが結果がよい
結論
- 結果重視ならGPT-4
- コストを気にするならGPT-3.5
- トークンのカウントは以下からできる
- https://platform.openai.com/tokenizer
- 出力トークンも費用が掛かる点に注意