# LLMRates.ai

> Real-time LLM API pricing intelligence. Compare costs, context windows, and capabilities across all major AI model providers.

## Key Pages

- [AI Model Pricing](https://llmrates.ai/models): Full comparison table — 1437 models from 21 providers, sortable and filterable
- [Provider Directory](https://llmrates.ai/providers): Individual provider pages with full model lineups and pricing history
- [Model Comparison](https://llmrates.ai/compare): Side-by-side comparison of up to 4 models
- [Open Pricing Dataset](https://llmrates.ai/open-llm-pricing-dataset): Free, daily-updated LLM pricing dataset (JSON + CSV, CC BY 4.0) — also on GitHub at https://github.com/llmrates/llm-pricing-dataset

## Providers Tracked

Anthropic, AWS Bedrock, Azure AI, Cohere, Cursor, DeepSeek, Google, Google Vertex AI, Groq, Meshs One, MiniMax, Mistral, Moonshot, OpenAI, OpenRouter, Perplexity, Qwen, SiliconFlow, Volcano Ark, xAI, Zhipu

## Cheapest Text Models (USD, standard tier, input price)

- Mistral OCR (25.05) (Google Vertex AI): input $0.0005 / output $0.0005 per 1M tokens
- DeepSeek-V4-Flash (Meshs One): input $0.0014 / output $0.0028 per 1M tokens
- Qwen3-235B-A22B (Meshs One): input $0.006 / output $0.01 per 1M tokens
- aisp-auto-route (Meshs One): input $0.01 / output $0.01 per 1M tokens
- inclusionAI: Ling-2.6-flash (OpenRouter): input $0.01 / output $0.03 per 1M tokens
- IBM: Granite 4.0 Micro (OpenRouter): input $0.017 / output $0.112 per 1M tokens
- Mistral: Mistral Nemo (OpenRouter): input $0.02 / output $0.04 per 1M tokens
- Nex AGI: Nex-N2-Mini (OpenRouter): input $0.025 / output $0.1 per 1M tokens
- Meta: Llama 3.2 1B Instruct (OpenRouter): input $0.027 / output $0.201 per 1M tokens
- LiquidAI: LFM2-24B-A2B (OpenRouter): input $0.03 / output $0.12 per 1M tokens
- OpenAI: gpt-oss-20b (OpenRouter): input $0.03 / output $0.13 per 1M tokens
- Amazon: Nova Micro 1.0 (OpenRouter): input $0.035 / output $0.14 per 1M tokens
- OpenAI: gpt-oss-120b (OpenRouter): input $0.037 / output $0.17 per 1M tokens
- Command R7B (Cohere): input $0.0375 / output $0.15 per 1M tokens
- Cohere: Command R7B (12-2024) (OpenRouter): input $0.0375 / output $0.15 per 1M tokens
- Gemma 3 4B (AWS Bedrock): input $0.04 / output $0.08 per 1M tokens
- Gemma 4 E2B (AWS Bedrock): input $0.04 / output $0.08 per 1M tokens
- Voxtral Mini 1.0 (AWS Bedrock): input $0.04 / output $0.04 per 1M tokens
- NVIDIA: Nemotron Nano 9B V2 (OpenRouter): input $0.04 / output $0.16 per 1M tokens
- Qwen: Qwen2.5 7B Instruct (OpenRouter): input $0.04 / output $0.1 per 1M tokens

## Largest Context Windows

- Meta: Llama 4 Scout (OpenRouter): 10M
- Auto Router (OpenRouter): 2M
- Pareto Code Router (OpenRouter): 2M
- xAI: Grok 4.20 (OpenRouter): 2M
- xAI: Grok 4.20 Multi-Agent (OpenRouter): 2M
- OpenAI: GPT-5.4 (OpenRouter): 1.1M
- OpenAI: GPT-5.4 Pro (OpenRouter): 1.1M
- OpenAI: GPT-5.5 (OpenRouter): 1.1M
- OpenAI: GPT-5.5 Pro (OpenRouter): 1.1M
- OpenAI: GPT-5.6 Luna (OpenRouter): 1.1M

## Data Notes

- Input price: cost per 1M tokens sent to the model (prompt + history)
- Output price: cost per 1M tokens generated by the model (response)
- Output tokens typically cost 2–5× more than input tokens
- Prices are scraped from official provider pages and updated at least weekly
- Batch pricing (where available) is ~50% of standard rates
- Prompt caching (Anthropic, Google) reduces repeated-prefix costs by 70–90%