GPT-Realtime-2Pricing & API costs
Reasoning model with tool use
- Input
- $4.00
- Output
- $24.00
- Cached input
- $0.40
Context window
128,000 tokens
Maximum output
32,000 tokens
Release date
Not published
API model ID
gpt-realtime-2Model overview
Understand the model before estimating your workload.
Use the model ID gpt-realtime-2 through the provider's direct API. The prices and limits on this page apply to that specific listing, with source data checked on 2026-10-12.
GPT-Realtime-2 has a published context window of 128,000 tokens. The maximum output for this channel is 32,000 tokens. Budget context space for your prompt, conversation history, retrieved material, and the response. Context capacity and output limits are different; your input and response must fit within the request limits. A larger window does not make those tokens free.
Pricing
Rates are grouped by usage type and available tier.
Audio
Per 1M tokens
Input
$32.00
Cached input
$0.40
Output
$64.00
Text
Per 1M tokens
Input
$4.00
Cached input
$0.40
Output
$24.00
Image
Per 1M tokens
Input
$5.00
Cached input
$0.50
Output
Not published
Batch pricing
Not published
How billing works
The details that turn a token rate into an actual bill.
Separate input from output
For GPT-Realtime-2, multiply input and output token counts by their respective rates, then divide by one million. Conversation history and retrieved context contribute to input usage each time they are sent.
Cache hits have their own rate
The listed cached-input price is $0.40 per million tokens. Reusing a prompt does not automatically qualify: use the provider's reported cached-token count, and account for any cache-write or storage charges.
Check discounts and rate tiers
No batch multiplier is applied to the examples on this page. Check the pricing source before assuming an asynchronous job receives a discount. Review the provider's pricing terms for workload-specific conditions.
Account for the complete workflow
Retries, follow-up calls, and agent tool loops increase request counts. Search, storage, and other tool charges can be billed separately from model inference. Use actual API usage logs to calibrate a budget before scaling traffic.
Capabilities and limits
What the API supports and where limits apply.
- Provider
- OpenAI
- Model type
- Audio
- Family
- gpt
- Access
- Direct API
- Modalities
- Text · Image · Audio
- Context window
- 128,000 tokens
- Max output
- 32,000 tokens
- Released
- Not published
- Deprecation date
- Not announced
Compare API channels
Check the model ID, rate, and limit for each access route.
The same model can have a different ID, price, or output limit depending on how you access it. These are the channels currently tracked in this catalog; they are not a complete list of every hosting provider.
Audio: $32.00 input · $0.40 cached input · $64.00 output / 1M tokens; Text: $4.00 input · $0.40 cached input · $24.00 output / 1M tokens; Image: $5.00 input · $0.50 cached input / 1M tokens
Sources
Official pages and the date this information was checked.
- Last checked
- 2026-10-12
- Source links
- 3
Next step
Estimate costs with this model.
Recommended for
Frequently asked questions
OpenAI pricing checked on 2026-10-12: Audio: $32.00 input · $0.40 cached input · $64.00 output / 1M tokens; Text: $4.00 input · $0.40 cached input · $24.00 output / 1M tokens; Image: $5.00 input · $0.50 cached input / 1M tokens. These rates apply to the listed API model ID.
GPT-Realtime-2 has a published context window of 128,000 tokens. The maximum output for this channel is 32,000 tokens. Budget context space for your prompt, conversation history, retrieved material, and the response. Context capacity and output limits are different; your input and response must fit within the request limits. A larger window does not make those tokens free.
The recorded status is Active. Use the exact model ID gpt-realtime-2 and confirm that your API account has access.
Recent updates
Added to the model directory.
View sourceModel details changed in the provider's documentation.
View source