AI / LLM
LLM API Pricing Guide 2025: GPT-4o, Claude, Gemini Compared
⚠️ Prices updated as of April 2025. LLM pricing changes frequently — verify at provider websites before billing clients.
Advertisement
As AI features become standard in software products, understanding LLM API costs is essential for developers and freelancers. Here's a comprehensive breakdown of all major models as of April 2025.
Current LLM API Pricing Table
| Model | Provider | Input (per 1K tok) | Output (per 1K tok) | Context | Best For |
|---|---|---|---|---|---|
| GPT-4o | OpenAI | $0.005 | $0.015 | 128K | Best all-round performance |
| GPT-4o mini | OpenAI | $0.00015 | $0.0006 | 128K | Budget OpenAI; production apps |
| o3-mini | OpenAI | $0.0011 | $0.0044 | 200K | Coding, reasoning tasks |
| Claude 3.5 Sonnet | Anthropic | $0.003 | $0.015 | 200K | Coding, writing, complex reasoning |
| Claude 3 Haiku | Anthropic | $0.00025 | $0.00125 | 200K | Fast, cheap Claude for classification |
| Gemini 1.5 Pro | $0.00125 | $0.005 | 2M ctx! | Very long documents/RAG | |
| Gemini 1.5 Flash | $0.000075 | $0.0003 | 1M ctx | Cheapest per token at scale | |
| Gemini 2.0 Flash | $0.0001 | $0.0004 | 1M ctx | Fast, affordable, latest Google | |
| Mistral Large | Mistral | $0.003 | $0.009 | 128K | European data regulatory needs |
Understanding Tokens
- ~1 token = 4 characters = 0.75 words (English)
- 750 words ≈ 1,000 tokens
- A typical ChatGPT conversation (1K input + 500 output) costs ~$0.007 with GPT-4o
- Output tokens typically cost 2–3× more than input tokens (generating is harder)
Which Model is Best for Different Use Cases?
| Use Case | Recommended Model | Why |
|---|---|---|
| Customer support chatbot (high volume) | Gemini 1.5 Flash or GPT-4o mini | Cheapest per token, fast latency |
| Code generation tool | Claude 3.5 Sonnet or o3-mini | Best coding performance |
| Long document analysis (>50K words) | Gemini 1.5 Pro (2M ctx) | Only model with 2M context window |
| Content generation (marketing, SEO) | GPT-4o or Claude 3.5 Sonnet | Best writing quality |
| Classification/labelling at scale | Claude 3 Haiku or GPT-4o mini | Fast, cheap, sufficient quality |
| Reasoning/math heavy tasks | o3-mini or GPT-4o | Best reasoning benchmarks |
GST on LLM APIs for Indian Businesses
OpenAI, Anthropic, and Google are foreign service providers. Indian GST-registered businesses must pay 18% IGST under Reverse Charge Mechanism (RCM) on these API charges:
- You pay the dollar invoice amount (in USD)
- Self-assess 18% IGST and pay it when filing GSTR-3B
- Input Tax Credit can be claimed on this RCM GST (offsets your output GST)
- Non-GST-registered individuals: RCM does not apply, no GST compliance needed