I’m using the Web Search tool and noticed the following pricing details:
Web search (all models): $10 / 1k calls + search content tokens billed at model rates
Web search preview (reasoning models, including gpt-5, o-series): $10 / 1k calls + search content tokens billed at model rates
Web search preview (non-reasoning models): $25 / 1k calls + search content tokens are free
I’m a bit confused — if I want to use the $10 / 1k calls tier, which model should I use (for example, gpt-4o, gpt-5, or o-series)? The documentation is not very clear about which models are included in each tier.
I can see your confusion the (all-models) part is the cause I think, I’ll ask the documentation team to take a look at that.
Essentially, all of the models that have reasoning (note: for gpt-5: use the “minimal” reasoning setting for web searches) will cost you $10 per 1k + search content tpkens, so, that’s GPT-5, o3, etc. and non reasoning, 4o, 4.1, etc at $25 per 1k + search and content toekns.
Also, ensure you are using the responses API for those calls.
Web search is available in the Responses API as the generally available version of the tool, web_search, as well as the earlier tool version, web_search_preview.
The web_search tool (i.e., general availability) will have search content tokens billed at model rates.
On the older web_search_preview, the content tokens are free for non-reasoning models like gpt-4o, gpt-4.1, etc., and users pay only for the respective search pricing per API call. For reasoning models like gpt-5, o3, and o4-mini, the content tokens are billed at model rates along with the respective search pricing per API call.
Hence to use the $10k / 1k calls pricing, you can use the web-search tool, regardless of the model you choose.
To avail the same pricing on the older web_search_preview tool, you’ll have to use it with a with a reasoning model that supports the tool.
Note: In both the cases you’ll be paying for the search content tokens at model rates.