impulse

GPT-4o Realtime Preview

Frontier labRealtimePreview
API id gpt-4o-realtime-preview-2024-12-17Released 17 Dec 2024 (Dec 2024)
Input / 1M
$5.00
at Azure OpenAI
Output / 1M
$20.00
at Azure OpenAI
Context window
16K
tokens
Max output
4K
tokens per reply
Hosts
1
3 listings
Released
Dec 2024
17 Dec 2024

Pricing

List prices at Azure OpenAI, in US dollars per million tokens.

TierInput / 1MOutput / 1M
Standard$5.00$20.00
Cached input
50% off input
$2.50

What it costs a month

At Azure OpenAI list prices, 30 days.

  • Chat assistant5,000 requests a day, 2K in / 0.5K out, cached$2,775
  • RAG over documents2,000 requests a day, 8K in / 0.6K out, cached$2,520
  • Coding agent1,000 requests a day, 40K in / 2K out, cached$5,100
  • Overnight batch20,000 requests a day, 4K in / 0.4K out$16,800
Use your own numbers

Where to run it

1 host serve GPT-4o Realtime Preview. The comparison is on blended price, 3 parts input to 1 part output.

HostIn / out per 1M
2024-12-17
gpt-4o-realtime-preview-2024-12-17
$5.00 / $20.00
Show 2 more listings: regions, snapshots and variants
Azure OpenAI
eu · 2024-12-17
gpt-4o-realtime-preview-2024-12-17
$5.50 / $22.00
+10% vs headline
us · 2024-12-17
gpt-4o-realtime-preview-2024-12-17
$5.50 / $22.00
+10% vs headline

Capabilities

Core

  • Streaming
  • Function calling
  • JSON mode
  • Structured outputs

Multimodal

  • Vision
  • Image input
  • Audio
  • Image generation
  • Video

Advanced

  • Reasoning
  • Tool calling
  • MCP
  • Prompt caching
  • Batch API
  • Embeddings
Takes in
AudioText
Produces
AudioText
Features
Audio inputAudio outputParallel tool calls

Cheaper alternatives

Current models that cost less and keep most of what this one can do: tools, vision, reasoning, structured output, caching and enough context.

Same model, cheaper host

No host sells it for less than Azure OpenAI.

Smaller OpenAI models

Other frontier labs

Open-weight picks

No cheaper open-weight match.

Call it

Using Azure OpenAI's API.

Python
from litellm import completion

response = completion(
    model="azure/gpt-4o-realtime-preview-2024-12-17",
    messages=[{"role": "user", "content": "Hello"}],
)
print(response.choices[0].message.content)

Quick answers

How much does GPT-4o Realtime Preview cost?

On Azure OpenAI, GPT-4o Realtime Preview costs $5.00 per million input tokens and $20.00 per million output tokens. Cached input costs $2.50 per million tokens.

Which APIs offer GPT-4o Realtime Preview?

GPT-4o Realtime Preview runs on 1 API: Azure OpenAI.

What is GPT-4o Realtime Preview's context window?

16K tokens, with up to 4K tokens in one reply.

Is GPT-4o Realtime Preview being retired?

No retirement date is scheduled for it on any API we track.

What are cheaper alternatives to GPT-4o Realtime Preview?

Models that cost less and keep most of what it can do: GPT Realtime 2.1 Mini on OpenAI (88% less), Gemini 3.8 Live on Google AI Studio (81% less) and Gemini 3.8 Live Extended Thinking on Google AI Studio (81% less). Savings are on a mix of 3 input tokens to 1 output token.

When was GPT-4o Realtime Preview released?

OpenAI released it on 17 Dec 2024.

What is GPT-4o Realtime Preview's API model id?

Use gpt-4o-realtime-preview-2024-12-17 with Azure OpenAI.

Sources

Raw catalog entry for gpt-4o-realtime-preview-2024-12-17 on Azure OpenAI
{
  "cache_read_input_audio_token_cost": 0.0000025,
  "cache_read_input_token_cost": 0.0000025,
  "input_cost_per_audio_token": 0.00004,
  "input_cost_per_token": 0.000005,
  "litellm_provider": "azure",
  "max_input_tokens": 16000,
  "max_output_tokens": 4096,
  "mode": "realtime",
  "output_cost_per_audio_token": 0.00008,
  "output_cost_per_token": 0.00002,
  "source": "https://azure.microsoft.com/en-us/pricing/details/cognitive-services/openai-service/",
  "supported_modalities": [
    "text",
    "audio"
  ],
  "supported_output_modalities": [
    "text",
    "audio"
  ],
  "supports_audio_input": true,
  "supports_audio_output": true,
  "supports_function_calling": true,
  "supports_parallel_function_calling": true,
  "supports_system_messages": true,
  "supports_tool_choice": true
}