AI prices: what does each model cost?
See what AI models cost to use, find the ones that are free right now and check which open models can run on your own computer.
Updated October 22, 2026. Sources: OpenRouter and Hugging Face.
- Cohere: North Mini Code (free)coherePrice per 1M tokens (in / out)Free1,000 chat answersFreeContext256K tokensOn your computer≈ 32 GB RAM
- Dots Studio: Dots3-Note Preview (free)dots-studioPrice per 1M tokens (in / out)Free1,000 chat answersFreeContext512K tokensOn your computerClosed model
- Google: Gemma 4 26B A4B (free)googlePrice per 1M tokens (in / out)Free1,000 chat answersFreeContext262K tokensOn your computer≈ 32 GB RAM
- Google: Gemma 4 31B (free)googlePrice per 1M tokens (in / out)Free1,000 chat answersFreeContext262K tokensOn your computer≈ 32 GB RAM
- Google: Lyria 3 Clip PreviewgooglePrice per 1M tokens (in / out)Free1,000 chat answersFreeContext1049K tokensOn your computerClosed model
- Google: Lyria 3 Pro PreviewgooglePrice per 1M tokens (in / out)Free1,000 chat answersFreeContext1049K tokensOn your computerClosed model
- LiquidAI: LFM2.5-2.6B (free)liquidPrice per 1M tokens (in / out)Free1,000 chat answersFreeContext66K tokensOn your computer≈ 8 GB RAM
- NVIDIA: Nemotron 3 Nano Omni (free)nvidiaPrice per 1M tokens (in / out)Free1,000 chat answersFreeContext256K tokensOn your computer≈ 32 GB RAM
- NVIDIA: Nemotron 3 Super (free)nvidiaPrice per 1M tokens (in / out)Free1,000 chat answersFreeContext262K tokensOn your computer≈ 128 GB RAM
- NVIDIA: Nemotron 3 Ultra (free)nvidiaPrice per 1M tokens (in / out)Free1,000 chat answersFreeContext1000K tokensOn your computerServer only
- NVIDIA: Nemotron 3.5 Content Safety (free)nvidiaPrice per 1M tokens (in / out)Free1,000 chat answersFreeContext128K tokensOn your computer≈ 8 GB RAM
- NVIDIA: Nemotron 3.5 Lightning (free)nvidiaPrice per 1M tokens (in / out)Free1,000 chat answersFreeContext1000K tokensOn your computer≈ 32 GB RAM
- NVIDIA: Nemotron 3.5 Super VL (free)nvidiaPrice per 1M tokens (in / out)Free1,000 chat answersFreeContext1000K tokensOn your computer–
- Poolside: Laguna S 2.1 (free)poolsidePrice per 1M tokens (in / out)Free1,000 chat answersFreeContext262K tokensOn your computer≈ 128 GB RAM
- Poolside: Laguna XS 2.1 (free)poolsidePrice per 1M tokens (in / out)Free1,000 chat answersFreeContext262K tokensOn your computer≈ 32 GB RAM
- Recursive: Flint (free)recursivePrice per 1M tokens (in / out)Free1,000 chat answersFreeContext262K tokensOn your computer≈ 32 GB RAM
- Thinking Machines: Inkling (free)thinkingmachinesPrice per 1M tokens (in / out)Free1,000 chat answersFreeContext1049K tokensOn your computerServer only
- Thinking Machines: Inkling Small (free)thinkingmachinesPrice per 1M tokens (in / out)Free1,000 chat answersFreeContext1049K tokensOn your computerServer only
- XingChen: Xing4.0 29B A4B (free)xingchen-agiPrice per 1M tokens (in / out)Free1,000 chat answersFreeContext262K tokensOn your computer≈ 32 GB RAM
1 million tokens is roughly 750,000 English words. A chat answer is counted as 500 tokens in and 500 out. Prices are API prices in US dollars via OpenRouter; apps such as ChatGPT or Claude have their own subscriptions.
RAM estimate for a 4-bit (Q4) version of the open model. Models with many experts still need memory for all parameters.
Free models often have usage limits and some providers may use your prompts for training. Check the terms before you send anything private.