GPT-4o Search Previewis a specialized model for web search in Chat Completions. It is trained to understand and execute web search queries.
GPT-4o Search Preview is a text-only model from OpenAI, offered as part of the GPT-4o family. It is accessible via OrcaRouter's API at the base URL https://api.orcarouter.ai/v1 under the model…
The model is well-suited for tasks that require reasoning over long text sequences, such as summarization, question-answering from large documents, and generating coherent responses in extended dialogues. Given its search preview label, it likely handles search result aggregation and contextual retrieval effectively. Its large context window enables analysis of entire reports, books, or codebases in a single prompt.
The primary limitation is the lack of multimodal support. It cannot process images, audio, or video inputs, which restricts its use in applications like visual question-answering or speech transcription. Additionally, no official benchmark scores or latency figures are available, making it difficult to compare performance on specific tasks. The model may not be optimized for tasks outside text-based reasoning.
For tasks with small context requirements (e.g., simple classification, short text generation), a cheaper and faster model such as GPT-3.5 Turbo may be more cost-effective. If multimodal inputs are needed, consider the standard GPT-4o model. For extremely high-volume text-only workloads, evaluate whether the zero-markup pricing on OrcaRouter justifies the per-token cost compared to other providers.
No. GPT-4o Search Preview accepts only text inputs. It does not support images, audio, or video. This is a key differentiator from the standard GPT-4o model, which is multimodal. Users requiring visual or auditory understanding should select a different model.
OpenAI has not published any benchmark scores specifically for the GPT-4o Search Preview variant. As it is part of the GPT-4o family, it may share performance characteristics with GPT-4o on language understanding tasks, but no official numbers are provided. Users are advised to conduct their own evaluations on representative datasets to determine suitability.
Latency and throughput figures are not publicly documented for GPT-4o Search Preview. As a variant of GPT-4o, it is likely to have similar response times to the standard GPT-4o model, but actual performance depends on factors like request concurrency, token count, and OrcaRouter infrastructure. Users should benchmark using their own workload.
Beyond the absence of multimodal support, the model lacks published benchmark scores and latency guarantees. Its search preview designation implies optimization for search-related tasks, but the extent of this optimization is not specified. Users should not assume it will outperform the standard GPT-4o on non-search tasks. Additionally, there is no guarantee of lower pricing or improved efficiency compared to other GPT-4o variants.
Pricing is $2.50 per 1 million input tokens and $10.00 per 1 million output tokens. OrcaRouter charges exactly the provider rate with zero markup. No additional fees are applied. Token usage is billed separately for input and output. These rates are consistent with OpenAI's standard GPT-4o pricing on their platform.
Compared to many other models, GPT-4o Search Preview is positioned at a premium per-token rate, typical of high-capability models. For example, GPT-3.5 Turbo is significantly cheaper. The zero-markup policy on OrcaRouter ensures users pay the same as direct OpenAI billing, but without any additional overhead. This can be advantageous for cost-sensitive applications where a large context window is required.
OrcaRouter does not publicly advertise any caching discounts or token savings programs specifically for GPT-4o Search Preview. Users are billed based on the total number of input and output tokens consumed. To manage costs, consider optimizing prompt lengths and limiting output tokens. Batch processing may improve efficiency but is not mentioned in the available documentation.
The model is accessed through OrcaRouter's OpenAI-compatible API. Set the base URL to https://api.orcarouter.ai/v1 and include the model ID "openai/gpt-4o-search-preview" in your request. Use standard OpenAI client libraries (e.g., Python openai package) with your OrcaRouter API key. Example: openai.ChatCompletion.create(model="openai/gpt-4o-search-preview", messages=[...]).
Since the API is OpenAI-compatible, standard parameters are available: model, messages, max_tokens, temperature, top_p, frequency_penalty, presence_penalty, stop, stream, and others. The max_tokens parameter must not exceed the 16,384 token output limit. The model does not support multimodal inputs, so messages should only contain text content.
To switch from another GPT-4o variant (e.g., "gpt-4o" or "gpt-4o-mini"), simply change the model ID in your API call to "openai/gpt-4o-search-preview". Ensure that your requests contain only text content, as this variant does not accept images or audio. Adjust the max_tokens parameter if needed. No other code changes are required.
Authentication is handled via API keys. You need an OrcaRouter API key, which is passed in the request header (typically as "Authorization: Bearer <your_key>"). Obtain a key from the OrcaRouter platform. The same authentication is used for all models accessible through the gateway.
GPT-4o Search Preview is text-only, while standard GPT-4o supports text, images, and audio. The search preview variant may have underlying optimizations for search-related queries, but those are not detailed publicly. Pricing is identical: $2.50/M input and $10/M output. Both use a 128k context window and 16k max output. The main trade-off is modality support versus potentially specialized search performance.
GPT-4 Turbo (128k context, text+image) is a predecessor to GPT-4o. GPT-4o Search Preview is newer and part of the GPT-4o family, which generally offers improved reasoning and efficiency over GPT-4 Turbo. However, multimodal support in GPT-4 Turbo is limited to images, whereas GPT-4o is fully multimodal. Search Preview removes that advantage. Pricing for GPT-4 Turbo is typically $10/M input and $30/M output, making Search Preview cheaper (if using the given rates). Note that GPT-4 Turbo pricing is not provided here; this is general knowledge.
Other search-focused models (e.g., dedicated retrieval-augmented generation systems) may offer built-in web search or file retrieval. GPT-4o Search Preview does not include its own search functionality; it is a language model that can process search results provided in the prompt. If you need external retrieval, combine this model with a search API. Its large context window makes it suitable for aggregating many search results in a single request.
The primary trade-off is between modality support and potential search-task optimization. You gain a large context window and zero-markup pricing, but lose the ability to process images or audio. On known benchmarks, performance may be equivalent to GPT-4o, but no specific scores are available. For text-only tasks that benefit from a 128k context, this model is a cost-effective choice within the GPT-4o family.
OpenAI-compatible — keep the SDK you already use
https://api.orcarouter.ai/v1from openai import OpenAI
client = OpenAI(
base_url="https://api.orcarouter.ai/v1",
api_key="$ORCAROUTER_API_KEY",
)
response = client.chat.completions.create(
model="openai/gpt-4o-search-preview",
messages=[{"role": "user", "content": "Hello"}],
)
print(response.choices[0].message.content)max_tokensresponse_formatstreamstructured_outputsweb_search_options| Input / 1M tokens | $2.50 |
| Output / 1M tokens | $10.00 |
| Currency | USD |
Estimate based on list price
Estimate only — actual token counts depend on the provider's tokenizer.
GET /api/public/models/openai/gpt-4o-search-previewOpen @misc{orcarouter_gpt_4o_search_preview,
title = {GPT-4o Search Preview API},
author = {OpenAI},
year = {2025},
howpublished = {OrcaRouter},
url = {https://www.orcarouter.ai/models/openai/gpt-4o-search-preview}
}OpenAI. (2025). GPT-4o Search Preview API. OrcaRouter. https://www.orcarouter.ai/models/openai/gpt-4o-search-preview