How much does the OpenAI API cost?

OpenAI API pricing is per million tokens (MTok). GPT-4o: $2.50 input / $10 output per MTok. GPT-4o mini: $0.15 input / $0.60 output per MTok. o3: $10 input / $40 output per MTok. o1: $15 input / $60 output per MTok. GPT-3.5 Turbo: $0.50 input / $1.50 output per MTok. Most production apps use GPT-4o or GPT-4o mini for the best cost-quality balance.

What is GPT-4o and how much does it cost?

GPT-4o (omni) is OpenAI's flagship model combining text, vision, and audio capabilities. It costs $2.50 per million input tokens and $10 per million output tokens. With a 128K context window, it's 2x cheaper than GPT-4 Turbo while being faster. GPT-4o mini at $0.15/$0.60 per MTok is the budget option for simpler tasks.

What are the o3 and o1 reasoning models and their pricing?

o3 is OpenAI's most powerful reasoning model, designed for complex math, coding, and scientific tasks. It costs $10 input / $40 output per MTok. o1 is the previous generation reasoning model at $15 input / $60 output per MTok. o3-mini is a faster, cheaper reasoning option at $1.10 input / $4.40 output per MTok. These are significantly more expensive than GPT-4o but much better at multi-step reasoning.

Does OpenAI offer a free tier for the API?

OpenAI does not offer a permanent free tier for the API. New accounts previously received $5 in free credits, though this is subject to change. ChatGPT (the consumer product) has a free tier but it does not include API access. For API access you must add a payment method and credits to your account.

Has OpenAI changed API prices recently?

Yes, OpenAI has repeatedly cut API prices. GPT-4o launched in 2024 at half the price of GPT-4 Turbo. GPT-4o mini replaced GPT-3.5 Turbo as the budget option at significantly lower cost. OpenAI also introduced Batch API pricing (50% discount for async requests) in 2024. Prices generally trend downward as new models and efficiency improvements arrive.

What is the OpenAI Batch API and how much does it save?

The OpenAI Batch API processes requests asynchronously with a 24-hour turnaround time at 50% off regular pricing. GPT-4o batch: $1.25 input / $5.00 output per MTok. GPT-4o mini batch: $0.075 input / $0.30 output per MTok. Ideal for data classification, embedding generation, document processing, and other high-volume offline tasks.

OpenAI API Pricing 2026

Complete pricing for GPT-4o, o3, o1, and GPT-3.5 Turbo — with real cost scenarios for apps, chatbots, coding assistants, and AI agents.

OpenAI API Pricing by Model

OpenAI charges per million tokens (MTok). Input tokens (your prompt) and output tokens (the response) are priced separately. Output tokens are typically 3-4x more expensive.

Model	Input (per MTok)	Output (per MTok)	Context	Best For
GPT-4o Popular	$2.50	$10.00	128K	Most production AI features, vision, complex tasks
GPT-4o mini Budget	$0.15	$0.60	128K	High-volume tasks, classification, simple Q&A
o3 Reasoning	$10.00	$40.00	200K	Complex math, coding, scientific reasoning
o3-mini	$1.10	$4.40	200K	Faster reasoning, coding tasks, cost-efficient
o1	$15.00	$60.00	200K	Deep reasoning, research-grade analysis
o1-mini	$3.00	$12.00	128K	STEM reasoning, coding at lower cost
GPT-4 Turbo	$10.00	$30.00	128K	Legacy workflows still on GPT-4
GPT-3.5 Turbo Legacy	$0.50	$1.50	16K	Simple tasks, legacy integrations

Cost tip: For most new projects, choose GPT-4o mini (complex routing, high volume) or GPT-4o (quality-sensitive features). GPT-3.5 Turbo costs more per token than GPT-4o mini and has less capability — migrate if you're still on it.

Batch API Pricing (50% Off)

The OpenAI Batch API processes requests asynchronously with results in under 24 hours at half the standard price. Ideal for large-scale processing jobs, dataset analysis, embeddings generation, and classification tasks.

Model	Batch Input (per MTok)	Batch Output (per MTok)	Savings vs Standard
GPT-4o (Batch)	$1.25	$5.00	50% off
GPT-4o mini (Batch)	$0.075	$0.30	50% off
o3-mini (Batch)	$0.55	$2.20	50% off
GPT-3.5 Turbo (Batch)	$0.25	$0.75	50% off

Embeddings & Other API Pricing

Model / Feature	Price	Notes
text-embedding-3-small	$0.02/MTok	Best for search, RAG, semantic similarity
text-embedding-3-large	$0.13/MTok	Higher accuracy, larger dimensions
text-embedding-ada-002	$0.10/MTok	Legacy model, use 3-small instead
Whisper (Speech-to-Text)	$0.006/min	Audio transcription and translation
TTS (Text-to-Speech)	$15/MTok (chars)	6 voices, HD option at $30/MTok
DALL·E 3	$0.04–$0.12/image	Varies by quality (standard/HD) and size

API Cost Calculator

Model

Requests per month

Avg input tokens per request

Avg output tokens per request

Estimated monthly API cost $32.50

Real Cost Scenarios

Customer Support Chatbot (GPT-4o mini) — 50K conversations/month

50K conversations × 800 input tokens avg 40M input tokens

50K conversations × 300 output tokens avg 15M output tokens

GPT-4o mini: $0.15 input + $0.60 output per MTok —

Total monthly API cost ~$15.00

AI Coding Assistant (GPT-4o) — 1K developer users, 20 completions/day each

20M completions/month × 1,500 input tokens 30B input tokens

20M completions × 500 output tokens 10B output tokens

GPT-4o: $2.50 input + $10 output per MTok —

Total monthly API cost ~$175,000

Document Summarization Pipeline (Batch API) — 500K docs/month

500K docs × 2,000 input tokens (avg doc length) 1B input tokens

500K docs × 300 output tokens (summary) 150M output tokens

GPT-4o mini Batch: $0.075 input + $0.30 output per MTok —

Total monthly API cost ~$120.00

RAG Application with Embeddings (text-embedding-3-small)

Index 100K documents × 1,000 tokens each (one-time) 100M tokens = $2.00

50K queries/month × 200 tokens for embedding 10M tokens = $0.20/mo

50K GPT-4o mini responses (retrieval + generation) ~$7.50/mo

Total monthly API cost (after initial indexing) ~$7.70/mo

Which OpenAI Model Should You Use?

Use Case	Recommended Model	Why
High-volume chatbot, Q&A, extraction	GPT-4o mini	17x cheaper than GPT-4o, handles most tasks well
Complex reasoning, multi-step agents	GPT-4o	Best capability-to-cost ratio for demanding tasks
Math, science, competitive coding	o3-mini	Purpose-built for reasoning, much cheaper than o3
Research-grade analysis, PhD-level tasks	o3	Highest capability, justify cost for precision-critical work
Large-scale offline batch processing	GPT-4o mini (Batch)	50% cheaper with async API — best for ETL, classification
Semantic search, RAG retrieval	text-embedding-3-small	Best accuracy per dollar, 5x cheaper than ada-002
Voice apps, transcription	Whisper + TTS	$0.006/min audio + $15/MTok for speech synthesis

7 Ways to Cut OpenAI API Costs

Downgrade to GPT-4o mini for most tasks — it's 17x cheaper and handles 80% of use cases. Only escalate to GPT-4o when quality matters.
Use Batch API for async workloads — 50% off for jobs that don't need real-time responses. Document processing, embeddings, and classification are ideal.
Shorten your system prompt — every request includes the full system prompt. A 2,000-token system prompt on 100K requests = 200M extra tokens = $50 extra/month on GPT-4o mini.
Cache common completions — OpenAI's Prompt Caching gives 50% off on cached input tokens for repeated prompt prefixes. Automatically applied for prompts over 1,024 tokens.
Set max_tokens limits — cap output length. Verbose responses cost 4x more than inputs; constraining output is the fastest way to reduce costs.
Migrate off GPT-3.5 Turbo — GPT-4o mini is cheaper ($0.15 vs $0.50 per MTok input) and significantly more capable. No reason to stay on GPT-3.5.
Use streaming for UX, not for cost — streaming doesn't reduce token usage. Use it to improve perceived latency, not to save money.

OpenAI API Price History

OpenAI has consistently reduced prices as efficiency improves. Prices have dropped by 95%+ since GPT-4 launched in 2023.

Date	Model / Change	Impact
Mar 2023	GPT-4 launches	$30/$60 per MTok — most expensive model at launch
Nov 2023	GPT-4 Turbo launches	$10/$30 per MTok — 67% cheaper than original GPT-4
Jan 2024	GPT-3.5 Turbo price cut	Input cut from $1.00 to $0.50 per MTok (50% off)
May 2024	GPT-4o launches	$2.50/$10 per MTok — 75% cheaper than GPT-4 Turbo
Jul 2024	GPT-4o mini launches	$0.15/$0.60 — replaces GPT-3.5 Turbo at lower price with better capability
Sep 2024	o1 launches	$15/$60 per MTok — premium reasoning model, new category
Apr 2025	o3 launches	$10/$40 per MTok — most capable model, better value than o1
Ongoing	Prompt Caching (all models)	50% off on cached input prefix — automatic discount for long prompts

OpenAI API vs Claude API — Price Comparison

Both OpenAI and Anthropic offer competitive API pricing. Here's a direct comparison of flagship models:

Model	Input (per MTok)	Output (per MTok)	Context
GPT-4o (OpenAI)	$2.50	$10.00	128K
Claude 3.5 Sonnet (Anthropic)	$3.00	$15.00	200K
GPT-4o mini (OpenAI)	$0.15	$0.60	128K
Claude 3.5 Haiku (Anthropic)	$0.80	$4.00	200K
o3 (OpenAI)	$10.00	$40.00	200K
Claude 3 Opus (Anthropic)	$15.00	$75.00	200K

For a full comparison including benchmark performance and real-world use cases, see the OpenAI vs Claude API pricing comparison.

Get Alerted When OpenAI Changes API Prices

OpenAI has changed prices 8+ times since 2023. Set up instant alerts so you're never caught off guard during budget planning.

Set Up Price Alerts — Free Free API Access