GPT-5.4 API Pricing Calculator & Benchmarks

Model Details

Released

Mar 5, 2026

Knowledge Cutoff

Aug 31, 2025

Context Length

1,050,000

Max Output

128,000

Modalities

Text Image File → Text

Where is GPT-5.4 a perfect fit?

gpt-5.4 is built for real-world professional work and autonomous task execution, not just chat. Ideal for:
- Agentic workflows that operate software (spreadsheets, docs, tools)
- Complex multi-step tasks involving planning, execution, and iteration
- Financial modeling, legal analysis, and enterprise decision systems
- Coding, debugging, and full-stack development workflows
- Long-context document analysis (reports, datasets, knowledge bases)
- Productivity automation across apps (Excel, presentations, workflows)

Quick Model Estimate

Tokens Characters Words

Model

Provider

Modality

Service Tier

Context Size

Input Tokens

(USD 2.5000 per 1M tokens)

Output Tokens

(USD 15.0000 per 1M tokens)

No. of API Calls

Your GPT-5.4 Cost Estimate

💰 Total Cost

—

for 1000 input + 1000 output tokens

📥 Input (1000 × $2.500000) —

📤 Output (1000 × $15.000000) —

Cost Breakdown

📥 Input 📤 Output

Prices are watched for changes and checked against the provider's own page.

Pricing

Provider ↕	Modality ↕	Service Tier ↕	Input Price (per 1M tokens) ↕	Output Price (per 1M tokens) ↕	Cached Input (per 1M tokens) ↕	Context Size ↕	View
OpenAI	Text	Standard	$2.5000	$15.0000	$0.2500	Up to 272,000 tokens	→
OpenAI	Text	Standard	$5.0000	$22.5000	$0.5000	272,000+ tokens	→
OpenAI	Text	Batch	$1.2500	$7.5000	$0.1300	Up to 272,000 tokens	→
OpenAI	Text	Batch	$2.5000	$11.2500	$0.2500	272,000+ tokens	→
OpenAI	Text	Flex	$1.2500	$7.5000	$0.1300	Up to 272,000 tokens	→
OpenAI	Text	Flex	$2.5000	$11.2500	$0.2500	272,000+ tokens	→
OpenAI	Text	Priority	$5.0000	$30.0000	$0.5000	Up to 272,000 tokens	→

No longer offered by

The last prices each provider charged for GPT-5.4 before it stopped offering it. These are no longer on offer. Each row links to that provider's price history.

Provider ↕	Ended ↕	Modality ↕	Service Tier ↕	Input Price (per 1M tokens) ↕	Output Price (per 1M tokens) ↕	Cached Input (per 1M tokens) ↕	Context Size ↕	View
Perplexity	28th Sep 2026Provider stopped offering it	Text	Standard	$2.5000	$15.0000	$0.2500	Up to 272,000 tokens	→
Perplexity	28th Sep 2026Provider stopped offering it	Text	Standard	$5.0000	$22.5000	N/A	272,000+ tokens	→

Recent changes

When GPT-5.4 was newly listed on a provider, moved to legacy or deprecated, or retired.

28th Sep 2026 Discontinued
Perplexity stopped offering OpenAI's GPT-5.4.

See all changes, new releases and retirements →

Benchmarks

Scores from standardized evaluations by Artificial Analysis and Design Arena. Higher is better — the indices summarize overall ability, while the detailed scores break down performance on individual benchmarks.

Artificial Analysis

Higher is better · benchmarked by Artificial Analysis

Index scores by variant

Detailed scores

Variant

Intelligence Index

Overall intelligence across reasoning, knowledge & math evals

Better than 52% of 32 models

71.1

Coding Index

Coding ability across software-engineering evals

Better than 40% of 16 models

—

Agentic Index

Tool use & multi-step agent task performance

Reasoning

GPQA Diamond 92%

Graduate-level questions in biology, chemistry & physics

Humanity's Last Exam 43.7%

Expert-level questions across many academic domains

IFBench 73.9%

Precise following of detailed instructions

τ²-Bench Telecom 87.1%

Tool-using agent tasks in a telecom support setting

CritPt 23.4%

Unpublished physics research reasoning problems

AA-LCR 82%

Long-context reasoning across large inputs

Coding

Terminal-Bench Hard 57.6%

Complex command-line and terminal workflows

Knowledge

GDPval 37.4%

Economically valuable, real-world knowledge work

Omniscience Accuracy 50.8%

Breadth of factual knowledge across domains

Omniscience Non-Hallucination 8.3%

How reliably the model avoids fabricated answers

All variants

Variant	GPQA Diamond	Humanity's Last Exam	IFBench	τ²-Bench Telecom	AA-LCR	Terminal-Bench Hard	CritPt	GDPval	Omniscience Accuracy	Omniscience Non-Hallucination
GPT-5.4 (Xhigh)	92%	43.7%	73.9%	87.1%	82%	57.6%	23.4%	37.4%	50.8%	8.3%
GPT-5.4 (Low)	87.1%	30.8%	65.9%	74.6%	76.7%	43.2%	7.4%	—	47.8%	17.4%
GPT-5.4 (Non-reasoning)	74.8%	11.3%	48.4%	36%	58.3%	37.9%	0.6%	—	37.4%	15.3%

Design Arena

Elo rating by arena & category

Variant	Arena	Category	Elo	Win %	Percentile	Avg time (ms)
gpt-5.4	agents	androidnative	1,033	47.4%	3	—
gpt-5.4	agents	fullstack	1,017	40.9%	9	—
gpt-5.4	agents	godotgamedev	1,135	46.9%	31	—
gpt-5.4	agents	mobileapps	1,085	46.1%	6	—
gpt-5.4	agents	webapps	1,059	40.3%	9	—
galapagos	models	3d	1,097	42.3%	33	196,350
gpt-5.4-low	models	3d	1,062	38.9%	22	64,820
gpt-5.4-none	models	3d	1,045	37%	17	55,002
galapagos	models	asciiart	1,196	55.4%	69	68,553
gpt-5.4-low	models	asciiart	1,185	52.5%	61	34,395
gpt-5.4-none	models	asciiart	1,185	51.8%	61	29,453
lavender	models	codecategories	1,210	52.5%	62	163,587
galapagos	models	codecategories	1,208	52.7%	61	165,874
gpt-5.4-low	models	codecategories	1,178	49.2%	52	85,870
gpt-5.4-none	models	codecategories	1,176	49.2%	51	75,302
galapagos	models	dataviz	1,235	56.6%	70	156,953
gpt-5.4-low	models	dataviz	1,196	52.2%	56	70,128
gpt-5.4-none	models	dataviz	1,181	50.1%	50	57,559
galapagos	models	gamedev	1,237	57.6%	73	202,864
gpt-5.4-none	models	gamedev	1,193	53.6%	60	91,449
gpt-5.4-low	models	gamedev	1,191	52.8%	59	95,576
galapagos	models	svg	1,191	57.8%	69	65,033
gpt-5.4-low	models	svg	1,138	53.2%	50	23,361
gpt-5.4-none	models	svg	1,123	50.4%	39	20,402
galapagos	models	uicomponent	1,234	57.4%	69	116,591
gpt-5.4-low	models	uicomponent	1,210	63.6%	62	61,413
gpt-5.4-none	models	uicomponent	1,194	52.5%	56	54,764
lavender	models	website	1,218	52.5%	62	163,587
galapagos	models	website	1,215	52.7%	61	158,134
gpt-5.4-none	models	website	1,197	50.7%	59	79,410
gpt-5.4-low	models	website	1,196	50.5%	58	89,294

Compare GPT-5.4 with

Two models side by side: worked costs, every rate, and every benchmark figure both of them report.

Deepseek v4 Pro-0813 DeepSeek Full comparison → GLM 5 Z.ai Full comparison → GPT-5.4 Pro OpenAI Full comparison → GPT-5.4 Nano OpenAI Full comparison → GPT-5.4 Mini OpenAI Full comparison → Grok-4.7 xAI Full comparison →

Every comparison we publish →

Other Models in the GPT-5.4 Family

GPT-5.4 Mini

Active

Chat · 400,000 context

GPT-5.4 Nano

Legacy

Chat · 400,000 context

GPT-5.4 Pro

Active

Chat · 1,050,000 context

FAQs about GPT-5.4

How much does GPT-5.4 cost per 1M tokens?

GPT-5.4 costs between $1.25 and $5.00 per million input tokens, and between $7.50 and $30.00 per million output tokens. The range covers Standard, Batch, Flex and Priority pricing.

What does a typical workload cost with GPT-5.4?

1,000 requests of 2,000 input and 500 output tokens each — 2,000,000 input and 500,000 output tokens in total — costs $6.25 with GPT-5.4 at its lowest rates. Use the calculator on this page for your own volumes.

Where can I use GPT-5.4?

GPT-5.4 is available through OpenAI, from $1.25 per million input tokens.

Is there cheaper GPT-5.4 pricing than the standard rate?

Yes. Batch pricing for GPT-5.4 starts at $1.25 per million input tokens against $2.50 on the standard tier, about 50% lower.

What is GPT-5.4's context window?

GPT-5.4 has a context window of 1,050,000 tokens, and returns up to 128,000 tokens in a single response.

What input types does GPT-5.4 support?

GPT-5.4 accepts text, image and file input, and returns text.

What is GPT-5.4's knowledge cutoff?

GPT-5.4's training data runs to 31 August 2025. It has no built-in knowledge of events after that date.