OpenAI GPT-5.4 API Pricing (2026): Full Cost Table + Calculator

If you are searching for GPT-5.4 API pricing, here is the full, practical version: exact token prices, side-by-side model comparison, and a calculator-style method you can copy into your team workflow.

If you want to compare GPT, Claude, Gemini, and Grok output quality before choosing a pricing strategy, try AIMirrorHub: https://aimirrorhub.com.

Quick Answer

As of 2026, GPT-5.4 API pricing is $2.50 input / $0.25 cached input / $15.00 output per 1M tokens. The key is not just unit price—it is cost per accepted output after retries and edits.

Full OpenAI API Pricing Table (GPT Family)

ModelInput ($/1M)Cached Input ($/1M)Output ($/1M)
GPT-5.42.500.2515.00
GPT-5.4 Pro30.00180.00
GPT-5.21.750.17514.00
GPT-51.250.12510.00
GPT-5 mini0.250.0252.00
GPT-5 nano0.050.0050.40

Note: GPT-5.4 Pro currently has no publicly listed cached-input price on the official OpenAI model page, so this table keeps it as .

Important Pricing Notes

  • For GPT-5.4 / GPT-5.4 Pro with very long prompts (>272K input tokens), session pricing multipliers can apply.
  • Regional processing/data residency endpoints may include an additional uplift.
  • Tool usage (like search/computer-use flows) can add separate cost depending on your setup.

Token Cost Formula (Copy/Paste)

Use this exact formula for quick estimates:

Cost = (InputTokens / 1,000,000 × InputPrice) + (CachedInputTokens / 1,000,000 × CachedPrice) + (OutputTokens / 1,000,000 × OutputPrice)

If no cached input is used, set the cached term to zero.

Worked Examples

Example A: GPT-5.4 standard workflow

Assume one run uses:

  • Input: 20,000 tokens
  • Cached input: 10,000 tokens
  • Output: 6,000 tokens

Cost:

  • Input: 20,000/1,000,000 × 2.50 = $0.05
  • Cached input: 10,000/1,000,000 × 0.25 = $0.0025
  • Output: 6,000/1,000,000 × 15.00 = $0.09

Total ≈ $0.1425 / run

Example B: GPT-5 mini lightweight workflow

Assume one run uses:

  • Input: 20,000 tokens
  • Cached input: 10,000 tokens
  • Output: 6,000 tokens

Cost:

  • Input: 20,000/1,000,000 × 0.25 = $0.005
  • Cached input: 10,000/1,000,000 × 0.025 = $0.00025
  • Output: 6,000/1,000,000 × 2.00 = $0.012

Total ≈ $0.01725 / run

Why Unit Price Alone Misleads

Many teams optimize for the cheapest model and then lose money in hidden operations:

  • More retries
  • More manual editing
  • Slower QA cycles
  • Higher failure rates in multi-step tasks

The better KPI is:

Cost per accepted output = (Token cost + Human correction cost + QA cost) / accepted outputs

Which Model to Use for Which Budget

  • GPT-5.4 Pro: Maximum quality for very high-value, complex tasks.
  • GPT-5.4: Strong default for professional reasoning + tools.
  • GPT-5.2 / GPT-5: Mid-tier balance if you need lower unit cost with solid capability.
  • GPT-5 mini: Great for high-volume, well-defined workflows.
  • GPT-5 nano: Best for ultra-low-cost classification/summarization pipelines.

FAQ

What is GPT-5.4 API price per 1M tokens?

Input $2.50, cached input $0.25, output $15.00.

Is GPT-5.4 Pro much more expensive?

Yes. GPT-5.4 Pro is significantly higher-priced and usually reserved for high-value workloads.

Is GPT-5 mini pricing complete in this table?

Yes: input $0.25, cached input $0.025, output $2.00 per 1M tokens.

Is GPT-5 nano included too?

Yes: input $0.05, cached input $0.005, output $0.40 per 1M tokens.

Final Take

If your goal is accurate GPT-5.4 API pricing, use the full table above and budget by workflow, not by headline token price. For most teams, hybrid routing (mini/nano for simple tasks, 5.4 for complex tasks) gives the best cost-performance balance.

Need a single workspace to benchmark outputs before deciding spend? Start with AIMirrorHub: https://aimirrorhub.com.