LLM API Pricing Comparison (2026)
Input and output token costs for 37 models from OpenAI, Anthropic, Google, Meta, and others — from $0.035 to $150 per million tokens.
| # | provider | model | release_date | input_cost_per_m_tokens_usd | output_cost_per_m_tokens_usd | context_window_k | multimodal | open_weights | + |
|---|---|---|---|---|---|---|---|---|---|
| 1 | OpenAI | GPT-4o | 2024-05 | 2.5 | 10.0 | 128 | True | False | |
| 2 | OpenAI | GPT-4o mini | 2024-07 | 0.15 | 0.6 | 128 | True | False | |
| 3 | OpenAI | GPT-4.5 | 2025-02 | 75.0 | 150.0 | 128 | True | False | |
| 4 | OpenAI | o1 | 2024-09 | 15.0 | 60.0 | 200 | False | False | |
| 5 | OpenAI | o1 mini | 2024-09 | 1.1 | 4.4 | 128 | False | False | |
| 6 | OpenAI | o3 | 2025-04 | 10.0 | 40.0 | 200 | True | False | |
| 7 | OpenAI | o3 mini | 2025-01 | 1.1 | 4.4 | 200 | False | False | |
| 8 | OpenAI | o4 mini | 2025-04 | 1.1 | 4.4 | 200 | True | False | |
| 9 | Anthropic | Claude 3.5 Haiku | 2024-11 | 0.8 | 4.0 | 200 | True | False | |
| 10 | Anthropic | Claude 3.5 Sonnet | 2024-10 | 3.0 | 15.0 | 200 | True | False | |
| 11 | Anthropic | Claude 3.7 Sonnet | 2025-02 | 3.0 | 15.0 | 200 | True | False | |
| 12 | Anthropic | Claude 3 Opus | 2024-03 | 15.0 | 75.0 | 200 | True | False | |
| 13 | Anthropic | Claude 4 Sonnet | 2026-03 | 3.0 | 15.0 | 200 | True | False | |
| 14 | Gemini 1.5 Flash | 2024-05 | 0.075 | 0.3 | 1000 | True | False | ||
| 15 | Gemini 1.5 Pro | 2024-05 | 1.25 | 5.0 | 2000 | True | False | ||
| 16 | Gemini 2.0 Flash | 2025-01 | 0.1 | 0.4 | 1000 | True | False | ||
| 17 | Gemini 2.0 Pro | 2025-03 | 1.25 | 5.0 | 2000 | True | False | ||
| 18 | Gemini 2.5 Pro | 2025-03 | 1.25 | 10.0 | 1000 | True | False | ||
| 19 | Meta | Llama 3.1 8B | 2024-07 | 0.18 | 0.18 | 128 | False | True | |
| 20 | Meta | Llama 3.1 70B | 2024-07 | 0.88 | 0.88 | 128 | False | True | |
| 21 | Meta | Llama 3.1 405B | 2024-07 | 5.0 | 15.0 | 128 | False | True | |
| 22 | Meta | Llama 3.3 70B | 2024-12 | 0.59 | 0.79 | 128 | False | True | |
| 23 | Meta | Llama 4 Scout | 2025-04 | 0.17 | 0.17 | 10000 | True | True | |
| 24 | Meta | Llama 4 Maverick | 2025-04 | 0.19 | 0.19 | 1000 | True | True | |
| 25 | Mistral | Mistral Small | 2025-01 | 0.1 | 0.3 | 32 | False | True | |
| 26 | Mistral | Mistral Medium | 2025-01 | 0.4 | 2.0 | 128 | False | False | |
| 27 | Mistral | Mistral Large | 2024-11 | 2.0 | 6.0 | 128 | False | False | |
| 28 | DeepSeek | DeepSeek V3 | 2024-12 | 0.27 | 1.1 | 128 | False | True | |
| 29 | DeepSeek | DeepSeek R1 | 2025-01 | 0.55 | 2.19 | 128 | False | True | |
| 30 | DeepSeek | DeepSeek R2 | 2026-01 | 0.55 | 2.19 | 128 | False | True | |
| 31 | xAI | Grok 2 | 2024-08 | 2.0 | 10.0 | 131 | True | False | |
| 32 | xAI | Grok 3 | 2025-02 | 3.0 | 15.0 | 131 | True | False | |
| 33 | Cohere | Command R | 2024-03 | 0.15 | 0.6 | 128 | False | False | |
| 34 | Cohere | Command R+ | 2024-04 | 2.5 | 10.0 | 128 | False | False | |
| 35 | Amazon | Nova Micro | 2024-12 | 0.035 | 0.14 | 128 | False | False | |
| 36 | Amazon | Nova Lite | 2024-12 | 0.06 | 0.24 | 300 | True | False | |
| 37 | Amazon | Nova Pro | 2024-12 | 0.8 | 3.2 | 300 | True | False |
1–37 of 37
Rows per page:
1 / 1
LLM API Pricing Comparison (2026) — AI Analysis
Key Findings
- GPT-4.5 is the most expensive model at $75/$150 per M tokens (input/output) — 3x pricier than the next costliest, Claude 3 Opus ($15/$75)
- The budget tier ($0.03–$0.20 input) includes 10 models from 5 providers: Amazon, Google, Meta, Mistral, and OpenAI
- Anthropic charges a consistent 5x output-to-input ratio across all Claude models, while Meta open-weight models charge nearly the same for input and output (1:1 ratio)
- Google offers the largest context windows (up to 2M tokens) at mid-range prices, while Llama 4 Scout leads at 10M context for just $0.17/M input
- All 10 open-weight models are priced under $5/M input tokens; 8 of them cost under $1
Visualizations
Open-Weight vs Closed-Weight Pricing
Average Input Price by Provider
10 Cheapest Models by Input Cost
Most Expensive Models by Output Cost