Pricing data for AI models, typed and kept up to date daily from models.dev.
114 providers · 4255 models · Updated 2026-05-06
npm install ai-model-pricesimport { getPrice, findModel, findProvider, getModels } from 'ai-model-prices';
// By provider ID + model ID
getPrice('openai/gpt-4.1-mini', { noCacheInput: 1000, noCacheOutput: 500 });
// Two call signatures
getPrice('openai', 'gpt-4.1-mini', { noCacheInput: 1000, noCacheOutput: 500 });
getPrice('amazon-bedrock', 'claude-3-5-sonnet-20241022-v2:0', { noCacheInput: 500, noCacheOutput: 200 });
// Get a model
const model = getModel('openai', 'gpt-4.1-mini');
// Get a provider
const provider = getProvider('amazon-bedrock');Each provider has a lightweight adapter. Types are scoped to that provider's models only:
import { getPrice, getModel, getProvider, getModels } from 'ai-model-prices/openai';
import { getPrice as bedrockPrice } from 'ai-model-prices/amazon-bedrock';
// provider is pre-bound
const price = getPrice('gpt-4.1-mini', { noCacheInput: 1000, noCacheOutput: 500 });
const model = getModel('gpt-4.1-mini'); // typed to OpenAI models onlyIf bundle size matters, provider data is loaded on demand:
import { getPrice, getModels, filterModels } from 'ai-model-prices/async';
const models = await filterModels({ maxInputCost: 1, tool_call: true });Returns a PriceBreakdown, or null if the model has no pricing data.
// Three call signatures:
getPrice(model: Model, tokens: TokenUsage): PriceBreakdown | null
getPrice('providerId/modelId', tokens: TokenUsage): PriceBreakdown | null
getPrice('providerId', 'modelId', tokens: TokenUsage): PriceBreakdown | nullinterface TokenUsage {
noCacheInput?: number; // regular input tokens
noCacheOutput?: number; // output tokens
cacheInput?: number; // cache read tokens
cacheWrite?: number; // cache write tokens
reasoning?: number; // reasoning tokens (o1, etc.)
inputAudio?: number;
outputAudio?: number;
}
interface PriceBreakdown {
noCacheInput: number;
noCacheOutput: number;
cacheInput: number;
cacheWrite: number;
reasoning: number;
audio: number;
total: number; // USD
}Both arguments are typed to known IDs.
Look up a provider by ID.
Return everything from the bundled data.
filterModels({
provider: 'openai', // one or more provider IDs
tool_call: true,
reasoning: false,
maxInputCost: 5, // USD per 1M tokens
maxOutputCost: 20,
minContextWindow: 128_000,
inputModalities: ['image'],
})| Provider | ID |
|---|---|
| 302.AI | 302ai |
| Alibaba | alibaba |
| Scaleway | scaleway |
| NanoGPT | nano-gpt |
| Abacus | abacus |
| Perplexity Agent | perplexity-agent |
| SiliconFlow (China) | siliconflow-cn |
| submodel | submodel |
| MiniMax Coding Plan (minimax.io) | minimax-coding-plan |
| Perplexity | perplexity |
| DeepSeek | deepseek |
| Llama | llama |
| OpenRouter | openrouter |
| Fireworks AI | fireworks-ai |
| Kimi For Coding | kimi-for-coding |
| Moark | moark |
| OpenCode Go | opencode-go |
| IO.NET | io-net |
| Alibaba (China) | alibaba-cn |
| MiniMax Coding Plan (minimaxi.com) | minimax-cn-coding-plan |
| Jiekou.AI | jiekou |
| Bailing | bailing |
| iFlow | iflowcn |
| v0 | v0 |
| Hugging Face | huggingface |
| ZenMux | zenmux |
| Upstage | upstage |
| NovitaAI | novita-ai |
| Xiaomi Token Plan (China) | xiaomi-token-plan-cn |
| Weights & Biases | wandb |
| Chutes | chutes |
| DInference | dinference |
| Vivgrid | vivgrid |
| Deep Infra | deepinfra |
| Kilo Gateway | kilo |
| SAP AI Core | sap-ai-core |
| Morph | morph |
| Cloudflare AI Gateway | cloudflare-ai-gateway |
| GitHub Copilot | github-copilot |
| Mixlayer | mixlayer |
| Xiaomi Token Plan (Singapore) | xiaomi-token-plan-sgp |
| Z.AI | zai |
| OpenCode Zen | opencode |
| StepFun | stepfun |
| Nebius Token Factory | nebius |
| Poe | poe |
| Helicone | helicone |
| Z.AI Coding Plan | zai-coding-plan |
| Amazon Bedrock | amazon-bedrock |
| Baseten | baseten |
| FrogBot | frogbot |
| Zhipu AI Coding Plan | zhipuai-coding-plan |
| Alibaba Coding Plan | alibaba-coding-plan |
| Venice AI | venice |
| AIHubMix | aihubmix |
| Cerebras | cerebras |
| LMStudio | lmstudio |
| LucidQuery AI | lucidquery |
| Moonshot AI (China) | moonshotai-cn |
| Azure Cognitive Services | azure-cognitive-services |
| abliteration.ai | abliteration-ai |
| Wafer | wafer.ai |
| Cohere | cohere |
| CloudFerro Sherlock | cloudferro-sherlock |
| KUAE Cloud Coding Plan | kuae-cloud-coding-plan |
| xAI | xai |
| Meganova | meganova |
| Vertex (Anthropic) | google-vertex-anthropic |
| evroc | evroc |
| Synthetic | synthetic |
| Nvidia | nvidia |
| Inference | inference |
| Inception | inception |
| OpenAI | openai |
| Requesty | requesty |
| DigitalOcean | digitalocean |
| Vultr | vultr |
| Alibaba Coding Plan (China) | alibaba-coding-plan-cn |
| Mistral | mistral |
| OVHcloud AI Endpoints | ovhcloud |
| Friendli | friendli |
| Cortecs | cortecs |
| SiliconFlow | siliconflow |
| Vercel AI Gateway | vercel |
| MiniMax (minimax.io) | minimax |
| LLM Gateway | llmgateway |
| Vertex | google-vertex |
| Cloudflare Workers AI | cloudflare-workers-ai |
| Groq | groq |
| Azure | azure |
| FastRouter | fastrouter |
| STACKIT | stackit |
| Tencent Coding Plan (China) | tencent-coding-plan |
| Privatemode AI | privatemode-ai |
google |
|
| D.Run (China) | drun |
| Moonshot AI | moonshotai |
| Berget.AI | berget |
| GitHub Models | github-models |
| Neuralwatt | neuralwatt |
| Together AI | togetherai |
| QiHang | qihang-ai |
| Tencent TokenHub | tencent-tokenhub |
| Anthropic | anthropic |
| ModelScope | modelscope |
| HPC-AI | hpc-ai |
| GitLab Duo | gitlab |
| Xiaomi | xiaomi |
| Clarifai | clarifai |
| MiniMax (minimaxi.com) | minimax-cn |
| Regolo AI | regolo-ai |
| Xiaomi Token Plan (Europe) | xiaomi-token-plan-ams |
| Zhipu AI | zhipuai |
| Nova | nova |
MIT