# GPT-5.4 Pro - Core.Today AI API

> Maximum-capability tier of the GPT-5.4 family (dated snapshot). Uses more compute to think longer and deliver the most reliable answers on the hardest reasoning, coding and analysis tasks. Served only via the OpenAI Responses API. $30/$180 per million tokens (no cached-input discount); prompts above 272K input tokens are billed at 2x input / 1.5x output for the whole request ($60/$270 per million tokens).

- **Provider**: OpenAI
- **Model ID**: gpt-5.4-pro-2026-03-05
- **Category**: LLM
- **Credits**: 195 per 1K tokens (avg)
- **Speed**: Slow
- **Quality**: Ultra

## Model Specifications
- **Context Window**: 1M tokens
- **Max Output**: 128K tokens
- **Training Cutoff**: 2025-08
- **Supported Formats**: text, json, markdown
- **Compatible SDK**: OpenAI

### Capabilities
- Vision (image input)
- Function Calling
- Streaming
- JSON Mode
- System Prompt

### Token Pricing (per 1M tokens)
- **Input Tokens**: 55,740 credits ($37.16)
- **Output Tokens**: 334,440 credits ($223)

## Features
- Maximum-capability GPT-5.4 tier
- 1M token context window
- 128K max output tokens
- Extended thinking for hardest problems
- Responses API only (no chat completions)
- Native vision and function calling

## Use Cases
- Hardest multi-step reasoning problems
- High-stakes code review and architecture
- Scientific and mathematical analysis
- Long-document synthesis at maximum quality

## API Endpoint

Base URL: https://api.core.today
Endpoint: POST /llm/openai/v1/responses

## Authentication
Header: Authorization: Bearer YOUR_API_KEY

Note: LLM endpoints use OpenAI-compatible format with Authorization Bearer token.

## Input Parameters

### Required
- **model**: string (default: gpt-5.4-pro-2026-03-05) - Model identifier (dated snapshot id)
- **input**: string | array - Responses API input: a plain string or an array of input items (messages, tool results)

### Optional
- **instructions**: string - System-level instructions for the run
- **max_output_tokens**: integer - Maximum tokens in the response, including internal reasoning tokens (Responses API field; not max_tokens)
- **reasoning**: object - Reasoning options, e.g. { effort: "high" }
- **stream**: boolean (default: false) - Enable Server-Sent Events streaming

## Examples

### Hardest Reasoning
Maximum-effort analysis with GPT-5.4 Pro via the Responses API
```json
{
  "model": "gpt-5.4-pro-2026-03-05",
  "input": "Review this database migration plan for a zero-downtime rollout across 40 services and identify every failure mode with mitigations.",
  "max_output_tokens": 16000
}
```

## Response Format
```json
{
  "id": "chatcmpl-abc123",
  "object": "chat.completion",
  "choices": [
    {
      "index": 0,
      "message": {
        "role": "assistant",
        "content": "Response text here"
      },
      "finish_reason": "stop"
    }
  ],
  "usage": {
    "prompt_tokens": 100,
    "completion_tokens": 50,
    "total_tokens": 150
  }
}
```

## Tips
- Works with the OpenAI SDK - set base_url to https://ai.api.core.today/llm/openai/v1 and use client.responses.create()
- Dated snapshot id - pin this exact id for reproducible behavior (the alias gpt-5.4-pro resolves to it)
- Served only via the Responses API (POST /llm/openai/v1/responses) — chat completions is not supported for this model
- Reserve for the hardest problems - standard GPT-5.4 ($2.50/$15) is far cheaper for routine tasks
- Responses take longer due to extended internal thinking - budget generous max_output_tokens, reasoning tokens count toward the limit
- Keep prompts under 272K input tokens where possible — above that the whole request bills at 2x input / 1.5x output ($60/$270 per 1M)

## Documentation
https://platform.openai.com/docs/models