GPT-5 Search API: Powerful text and image model via OrcaRouter. 128K output tokens, competitive pricing.
The GPT-5 Search API is a model from OpenAI, accessible through OrcaRouter’s OpenAI-compatible API at https://api.orcarouter.ai/v1 with model ID "openai/gpt-5-search-api". It accepts text and image…
It can handle a wide variety of natural language and multimodal tasks, particularly those requiring search-enhanced reasoning. Examples include generating answers based on retrieved documents, analyzing images and providing textual summaries, converting long documents into structured formats, and creating extended content like reports or articles. Its high output limit makes it suitable for tasks where the response must be very long, such as generating full codebases or detailed analyses.
If your application does not require the high context length or multimodal input, smaller models like GPT-4o mini or GPT-3.5 Turbo may be more cost-effective. Tasks with short conversations, simple text classification, or low token counts can be handled by less expensive options. Additionally, if you do not need search-augmented capabilities, a standard GPT model might suffice. Evaluate your average output length and input complexity to decide whether the GPT-5 Search API’s premium is justified.
The model is designed to work with search results. While the specific implementation depends on your application, typical use involves feeding search snippets or retrieved documents as context. The model then generates answers grounded in that information. This approach improves factual accuracy and reduces hallucination. The model’s large output capacity allows it to synthesize information from many sources into a coherent, lengthy response.
Image input is valuable for tasks where visual information is essential. Examples include describing charts or diagrams, extracting text from images (with OCR), answering questions about photographs, or combining image analysis with text instructions. The model processes the image and text jointly, enabling it to answer queries like "What is shown in this image?" or "Describe the scene in detail." It can also handle multiple images in a single request.
As of the provided facts, no specific benchmark scores are publicly listed for this model. Performance characteristics are inferred from its design as a GPT-5-class model with search augmentation. It is expected to perform well on tasks requiring long-form generation, factual accuracy with retrieval, and multimodal reasoning. Users are encouraged to evaluate the model on their own datasets to measure relevance to their use case.
Latency depends on the length of input and output tokens, as well as the number of images. Processing images adds some overhead. Generally, generating 128K tokens will be slower than shorter outputs. OrcaRouter does not add extra latency beyond the provider’s response time. For production use, consider setting max_tokens to the minimum needed to balance speed and output quality. Streaming responses can reduce perceived latency.
Its primary strength is the combination of a very high output token limit (128K) with multimodal input (text and image). This makes it one of the few models that can accept images and produce extremely long text outputs. The search-augmented nature also helps in tasks where up-to-date or specific information must be retrieved. It is versatile enough for both creative generation and analytical tasks.
Limitations include the lack of publicly verified benchmark data, meaning performance on standard NLP tasks is not formally quantified. The model may require careful prompt engineering to handle multiple images effectively. Also, while the output is large, the context window for input is not specified; very long input sequences might exceed the model’s capacity. Pricing per token is higher than smaller models, so budget management is important.
Pricing is $1.25 per 1 million input tokens and $10.00 per 1 million output tokens. These are provider rates with zero markup from OrcaRouter. There are no additional platform fees or hidden costs. Charges are based on the number of tokens processed. Image tokens are billed according to OpenAI’s image pricing (typically per image based on resolution). Users pay only for what they use.
The output token cost is significantly higher than input cost (8x ratio). Therefore, applications that generate very long outputs can be expensive. To manage costs, consider limiting max_tokens to the minimum required, using shorter prompts, and caching common responses. For tasks with modest output lengths, cheaper models may be more economical. The zero markup ensures you are not paying extra through OrcaRouter.
No specific information about cached token discounts is provided for this model. OpenAI may offer prompt caching on some models, but it is not mentioned for GPT-5 Search API. Check OpenAI’s official documentation for the latest caching policies. OrcaRouter passes through billing exactly as the provider charges, so any provider-level discounts would apply accordingly.
Use the OpenAI-compatible endpoint at https://api.orcarouter.ai/v1/chat/completions. Set the model parameter to "openai/gpt-5-search-api". Include your OrcaRouter API key in the Authorization header. The request body follows the standard chat completions format with messages array. For image input, include the "image_url" content type. An example Python snippet: ```python import openai openai.api_base = "https://api.orcarouter.ai/v1" openai.api_key = "your_key" response = openai.ChatCompletion.create( model="openai/gpt-5-search-api", messages=[{"role": "user", "content": "Describe this image"}], max_tokens=1000) ```
Standard chat completion parameters are supported: messages, max_tokens, temperature, top_p, frequency_penalty, presence_penalty, stop, and stream. For image input, the content type must be set to "image_url" with a URL or base64 data. Use "text" for text content. The model does not support function calling or tool use in the current release. All parameters behave as documented by OpenAI.
Replace your existing model name (e.g., gpt-4o) with "openai/gpt-5-search-api" and change the base URL to OrcaRouter’s endpoint. If your application uses image inputs, ensure you format them correctly. No other changes are needed. You may adjust max_tokens to benefit from the higher limit. Test with a few requests initially to verify output quality.
No. OrcaRouter is a pass-through gateway that routes requests directly to OpenAI’s API with zero added latency or processing. There are no rate limits beyond those imposed by OpenAI. The service does not modify requests or responses. This ensures you get the same performance as calling OpenAI directly, with the convenience of a unified API for multiple providers.
GPT-5 Search API offers a higher output token limit (128K vs. GPT-4o’s 16K) and supports image input, whereas GPT-4o is text-only. The search-augmented design also differentiates it for retrieval tasks. Pricing is higher per token. If you need very long outputs or image understanding, GPT-5 Search API is preferable. For standard text tasks, GPT-4o may be more cost-effective.
Models like Claude 3 Opus and Gemini 1.5 Pro also offer large context windows (up to 200K tokens). However, GPT-5 Search API’s specific search capability and image support set it apart. Pricing is competitive, though exact comparisons depend on token usage. The zero markup through OrcaRouter makes it simpler to budget. Each model has unique strengths; choose based on your need for search integration vs. raw context size.
Yes, it is significantly more capable in terms of output length, multimodal input, and reasoning. However, it is also much more expensive. For simple tasks like short Q&A or classification, GPT-3.5 Turbo remains a cost-effective choice. The choice depends on the complexity of your task and whether you need the advanced features.
Data handling follows OpenAI’s privacy policy. OrcaRouter does not log or store your prompts or responses. By default, OpenAI may use API data to improve models unless you opt out. Check OpenAI’s data usage policy for details. For sensitive data, consider using API keys that are part of an enterprise agreement or use the model via OrcaRouter’s privacy-focused settings if available.
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-5-search-api",
messages=[{"role": "user", "content": "Hello"}],
)
print(response.choices[0].message.content)max_tokensresponse_formatservice_tierstopstreamstructured_outputstool_choicetoolstop_pweb_search_options| Input / 1M tokens | $1.25 |
| Output / 1M tokens | $10.00 |
| Cache read / 1M | $0.125 |
| Currency | USD |
Estimate based on list price
Estimate only — actual token counts depend on the provider's tokenizer.
GET /api/public/models/openai/gpt-5-search-apiOpen @misc{orcarouter_gpt_5_search_api,
title = {openai/gpt-5-search-api API},
author = {openai},
year = {n.d.},
howpublished = {OrcaRouter},
url = {https://www.orcarouter.ai/models/openai/gpt-5-search-api}
}openai. (n.d.). openai/gpt-5-search-api API. OrcaRouter. https://www.orcarouter.ai/models/openai/gpt-5-search-api