Advanced AI model from OpenAI with 100k token output, text/image input, and state-of-the-art 99.4% on MATH-500.
openai/gpt-5-chat-latest is the latest chat model from OpenAI, available through OrcaRouter's API. It accepts text and image inputs and can generate up to 100,000 tokens of output. The provider is…
The model can handle a wide range of tasks that require complex reasoning: solving advanced math problems (as evidenced by its MATH-500 score), writing and debugging code, conducting scientific analysis, and understanding multimodal inputs like diagrams, charts, and photographs. It is particularly strong in tasks that require step-by-step logical deduction and long-form output. For simpler tasks, a cheaper model may be more cost-effective.
Yes, the model accepts image inputs. This means it can analyze visual data such as scanned documents, graphs, medical images, and photographs. It integrates visual understanding with textual reasoning, allowing tasks like describing a chart in detail, extracting text from an image, or explaining a diagram. The output is still text; the model does not generate images.
The cost of openai/gpt-5-chat-latest is higher than many alternatives ($1.25 input / $10 output per 1M tokens). For straightforward question-answering, summarization, or basic content generation, models like GPT-4o-mini or Claude Haiku may provide adequate quality at lower cost. Reserve this model for tasks where the extra reasoning power measurably improves outcomes, such as advanced math, complex code, or high-stakes analysis.
Optimal use cases include proving mathematical theorems, solving competition-level math problems, generating long, syntactically complex code, analyzing scientific papers with figures, and performing multi-step logical reasoning. It is also suitable for educational tools that require accurate explanations and for research that demands low error rates. The 100k token output enables generation of lengthy reports or entire codebases in one call.
The headline benchmark score is 99.4 on MATH-500, which is a dataset of 500 challenging math problems covering algebra, geometry, number theory, and more. A score of 99.4 means the model answered 99.4% of problems correctly. This places it among the highest-performing models for mathematical reasoning. It demonstrates the model's ability to apply consistent logic and handle multi-step problem solving.
Only the MATH-500 score is provided in the available information. No other benchmark results are listed for this model. Users should evaluate the model on their own tasks to confirm performance. The MATH-500 score indicates strong math reasoning, but performance on other domains (e.g., coding benchmarks, language understanding) may vary. For comprehensive evaluation, we recommend testing with your specific use cases.
Speed and latency data are not publicly specified for this model. The provider, OpenAI, typically offers low-latency responses for their models, but actual speed depends on factors such as input length, output length, and current load. OrcaRouter does not throttle or modify the model's performance, so latency mirrors that of the provider. For real-time applications, consider testing the model's response time with your typical workload.
Pricing is $1.25 per 1 million input tokens and $10.00 per 1 million output tokens. This is the provider rate with zero markup from OrcaRouter. Input tokens include the entire prompt, including images (which are tokenized). Output tokens include the model's response. Costs scale linearly with token consumption, so longer outputs cost more. There are no additional fees for API calls or throughput.
No information about caching discounts or volume pricing is provided. OrcaRouter does not advertise any caching discounts for this model. For volume pricing, users may need to contact providers directly. The current pay-as-you-go rate is the only transparent pricing available. If you expect heavy usage, contact OrcaRouter or OpenAI for potential custom agreements.
Compared to previous models like GPT-4 Turbo, this model is more expensive per token. GPT-4 Turbo typically costs $10/1M input and $30/1M output, so openai/gpt-5-chat-latest has lower output cost but higher input cost relative to that. Compared to smaller models like GPT-4o-mini ($0.15 input, $0.60 output), it is significantly pricier. The trade-off is higher accuracy on difficult tasks.
Using openai/gpt-5-chat-latest is a trade-off between cost and accuracy. For tasks where a wrong answer is costly (e.g., financial analysis, medical reasoning), the higher price can be justified by the model's superior performance on benchmarks like MATH-500. For routine tasks, the cost may exceed the benefit. Always evaluate whether the improvement in output quality compensates for the increased token cost.
You can call openai/gpt-5-chat-latest using OrcaRouter's OpenAI-compatible API. Set the base URL to https://api.orcarouter.ai/v1 and specify the model as "openai/gpt-5-chat-latest". All standard OpenAI SDKs and HTTP clients work with this endpoint. You must include a valid OrcaRouter API key. Example using Python: openai.api_base = 'https://api.orcarouter.ai/v1'; openai.api_key = 'your-key'; response = openai.ChatCompletion.create(model='openai/gpt-5-chat-latest', messages=[...])
The model supports standard OpenAI Chat Completion parameters: temperature (0-2), top_p, frequency_penalty, presence_penalty, max_tokens (up to 100,000), stop sequences, stream (boolean for streaming), and logit_bias. Additional OpenAI parameters like response_format or tools may also be supported. Refer to OpenAI's documentation for details. OrcaRouter passes these parameters directly to the provider.
Migration is straightforward: change the model ID in your API call from your previous model (e.g., "gpt-4") to "openai/gpt-5-chat-latest". No other code changes are needed if you are already using OrcaRouter's OpenAI-compatible endpoint. The same parameters and message structure apply. For best results, test your prompts with this model, as it may respond differently to formatting or instructions.
Yes, the model supports streaming via the 'stream' parameter set to true. When streaming, the response is sent as a series of chunks with delta content. This is useful for applications that need real-time output, such as chat interfaces. OrcaRouter's API handles streaming exactly as OpenAI does, so your existing streaming code should work by simply changing the model ID.
While GPT-4o is also a multimodal model from OpenAI, openai/gpt-5-chat-latest is the newer and more advanced version, as indicated by the '5' in its name. The MATH-500 score of 99.4 significantly exceeds typical GPT-4o performance on that benchmark (which is around mid-90s). The output limit of 100k tokens is also larger than GPT-4o's typical 16k-32k. However, pricing is higher for both input and output tokens.
Comparisons are based on available benchmarks. Claude 3.5 Sonnet scores around 78 on MATH-500, so openai/gpt-5-chat-latest's 99.4 is substantially higher. However, Claude models may excel in other areas like safety and long-context recall. Pricing for Claude 3.5 Sonnet is $3/1M input and $15/1M output, making it cheaper. The choice depends on whether math accuracy is critical for your use case.
Using OrcaRouter offers zero markup on provider pricing, so the cost is identical to going directly to OpenAI. Additionally, OrcaRouter provides a unified API for multiple providers, simplifying key management and billing. If you already use OrcaRouter for other models, adding openai/gpt-5-chat-latest requires just a model ID change. OrcaRouter may also offer features like fallback routing, usage analytics, and monitoring.
OpenAI-compatible — keep the SDK you already use
https://api.orcarouter.ai/v1import os
from openai import OpenAI
client = OpenAI(
base_url="https://api.orcarouter.ai/v1",
api_key=os.environ["ORCAROUTER_API_KEY"],
)
response = client.chat.completions.create(
model="openai/gpt-5-chat-latest",
messages=[{"role": "user", "content": "Hello"}],
)
print(response.choices[0].message.content)frequency_penaltylogit_biaslogprobsmax_tokensparallel_tool_callspresence_penaltypredictionresponse_formatseedservice_tierstopstreamstructured_outputstemperaturetool_choicetoolstop_logprobstop_p| 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.
What developers are saying this week
GET /api/public/models/openai/gpt-5-chat-latestOpen @misc{orcarouter_gpt_5_chat_latest,
title = {openai/gpt-5-chat-latest API},
author = {openai},
year = {n.d.},
howpublished = {OrcaRouter},
url = {https://www.orcarouter.ai/models/openai/gpt-5-chat-latest}
}openai. (n.d.). openai/gpt-5-chat-latest API. OrcaRouter. https://www.orcarouter.ai/models/openai/gpt-5-chat-latest