🏢AI API Pricing by Provider 2026

AI API Pricing by Provider

All 18 AI model APIs grouped by company — input & output price per 1M tokens, sorted cheapest first within each provider. Compare OpenAI vs Anthropic vs Google vs DeepSeek and more.

OpenAIAnthropicGoogleDeepSeekAlibabaMetaMistralxAI

OpenAI

3 models

Widest model range ($0.10–$15/1M input). GPT-4.1 Mini is the go-to for production; o-series for heavy reasoning.

ModelInput /1MOutput /1M
GPT-5.6 Luna$1.00/1M tkns$6.00/1M tknsCompare →
GPT-5.6 Terra$2.50/1M tkns$15.00/1M tknsCompare →
GPT-5.6 Sol$5.00/1M tkns$30.00/1M tknsCompare →
Cheapest: GPT-5.6 Luna at $1.00/1M tknsMost capable: GPT-5.6 Sol at $5.00/1M tkns

Anthropic

3 models

Premium quality, strong instruction-following ($0.25–$15/1M). Claude Sonnet 4.6 is the price-performance sweet spot.

ModelInput /1MOutput /1M
Claude Sonnet 5$2.00/1M tkns$10.00/1M tknsCompare →
Claude Opus 4.8$5.00/1M tkns$25.00/1M tknsCompare →
Claude Fable 5$10.00/1M tkns$50.00/1M tknsCompare →
Cheapest: Claude Sonnet 5 at $2.00/1M tknsMost capable: Claude Fable 5 at $10.00/1M tkns

Google

3 models

Best context windows and value flashes ($0.07–$2.50/1M). Gemini 2.5 Flash at $0.30/1M is a standout deal.

ModelInput /1MOutput /1M
Gemini 3.1 Flash-Lite$0.300/1M tkns$2.50/1M tknsCompare →
Gemini 2.5 Pro$1.25/1M tkns$10.00/1M tknsCompare →
Gemini 3.5 Flash$1.50/1M tkns$7.50/1M tknsCompare →
Cheapest: Gemini 3.1 Flash-Lite at $0.300/1M tknsMost capable: Gemini 3.5 Flash at $1.50/1M tkns

DeepSeek

2 models

Cheapest frontier models on the market ($0.07–$0.27/1M). MoE architecture delivers GPT-4-tier quality at 10× less.

ModelInput /1MOutput /1M
DeepSeek V4 Flash$0.140/1M tkns$0.280/1M tknsCompare →
DeepSeek V4 Pro$0.435/1M tkns$0.870/1M tknsCompare →
Cheapest: DeepSeek V4 Flash at $0.140/1M tknsMost capable: DeepSeek V4 Pro at $0.435/1M tkns

Alibaba

2 models

Qwen 2.5 72B offers strong multilingual performance at competitive pricing.

ModelInput /1MOutput /1M
Qwen3.7 Plus$0.320/1M tkns$1.28/1M tknsCompare →
Qwen3.7 Max$1.47/1M tkns$4.42/1M tknsCompare →
Cheapest: Qwen3.7 Plus at $0.320/1M tknsMost capable: Qwen3.7 Max at $1.47/1M tkns

Meta

2 models

Open-weight Llama models available via API. Llama 4 Scout at $0.17/1M is Meta's cost leader.

ModelInput /1MOutput /1M
Llama 4 Scout$0.100/1M tkns$0.300/1M tknsCompare →
Llama 4 Maverick$0.200/1M tkns$0.800/1M tknsCompare →
Cheapest: Llama 4 Scout at $0.100/1M tknsMost capable: Llama 4 Maverick at $0.200/1M tkns

Mistral

2 models

European AI lab with strong code and multilingual models ($0.10–$2/1M). Codestral is purpose-built for code.

ModelInput /1MOutput /1M
Mistral Large$0.500/1M tkns$1.50/1M tknsCompare →
Mistral Medium 3.5$1.50/1M tkns$7.50/1M tknsCompare →
Cheapest: Mistral Large at $0.500/1M tknsMost capable: Mistral Medium 3.5 at $1.50/1M tkns

xAI

1 model

Grok 3 and Grok 3 Mini from Elon Musk's AI lab. Competitive on reasoning, real-time data access.

ModelInput /1MOutput /1M
Grok 4.5$2.00/1M tkns$6.00/1M tkns
Cheapest: Grok 4.5 at $2.00/1M tkns

More AI Pricing Resources

🤖
Full AI Model Price Table
All models sorted by input price
🎯
AI Pricing by Use Case
Best model per workload and budget
📊
LLM API Pricing Guide
How token pricing works, explained
🔧
Best AI for Coding
Codestral, GPT-4.1, Claude Sonnet compared