ai-model-prices

Pricing data for AI models, typed and kept up to date daily from models.dev.

139 providers · 4931 models · Updated 2026-06-26

Installation

npm install ai-model-prices

Usage

Main package

import { 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');

Provider adapters

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 only

Async (lazy-loaded)

If 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 });

API

getPrice(model, tokens)

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 | null

interface 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
}

getModel(providerId, modelId)

Both arguments are typed to known IDs.

getProvider(id)

Look up a provider by ID.

getModels() / getProviders() / getModelsByProvider(providerId)

Return everything from the bundled data.

filterModels(filter)

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'],
})

Supported Providers

Provider	ID
Requesty	requesty
Qiniu	qiniu-ai
Alibaba (China)	alibaba-cn
Regolo AI	regolo-ai
STACKIT	stackit
Vercel AI Gateway	vercel
submodel	submodel
Hugging Face	huggingface
MiniMax Token Plan (minimax.io)	minimax-coding-plan
NovitaAI	novita-ai
xAI	xai
Privatemode AI	privatemode-ai
D.Run (China)	drun
Alibaba Token Plan (China)	alibaba-token-plan-cn
Moonshot AI	moonshotai
Fireworks AI	fireworks-ai
Vultr	vultr
302.AI	302ai
Zhipu AI	zhipuai
Cortecs	cortecs
Nebius Token Factory	nebius
Auriko	auriko
StepFun AI	stepfun-ai
Vivgrid	vivgrid
Mistral	mistral
Cloudflare Workers AI	cloudflare-workers-ai
Bailing	bailing
Google	google
OpenCode Go	opencode-go
DigitalOcean	digitalocean
Venice AI	venice
LMStudio	lmstudio
Poolside	poolside
ZenMux	zenmux
OpenAI	openai
Berget.AI	berget
GitHub Models	github-models
Neuralwatt	neuralwatt
SiliconFlow (China)	siliconflow-cn
Merge Gateway	merge-gateway
QiHang	qihang-ai
Xiaomi Token Plan (Europe)	xiaomi-token-plan-ams
ModelScope	modelscope
Groq	groq
Mixlayer	mixlayer
OrcaRouter	orcarouter
Helicone	helicone
Z.AI	zai
NEAR AI Cloud	nearai
LLM Gateway	llmgateway
Alibaba Coding Plan (China)	alibaba-coding-plan-cn
Abacus	abacus
CloudFerro Sherlock	cloudferro-sherlock
Cloudflare AI Gateway	cloudflare-ai-gateway
Moonshot AI (China)	moonshotai-cn
Morph	morph
Deep Infra	deepinfra
Vertex (Anthropic)	google-vertex-anthropic
v0	v0
Azure	azure
Cerebras	cerebras
Z.AI Coding Plan	zai-coding-plan
Nvidia	nvidia
evroc	evroc
Xiaomi	xiaomi
Inception	inception
Anthropic	anthropic
Tencent Coding Plan (China)	tencent-coding-plan
FreeModel	freemodel
SAP AI Core	sap-ai-core
OpenCode Zen	opencode
Inference	inference
Inceptron	inceptron
Llama	llama
LLMTR	llmtr
Cohere	cohere
StepFun	stepfun
HPC-AI	hpc-ai
MiniMax (minimaxi.com)	minimax-cn
Alibaba Coding Plan	alibaba-coding-plan
Poe	poe
Kimi For Coding	kimi-for-coding
DInference	dinference
Perplexity Agent	perplexity-agent
SiliconFlow	siliconflow
Umans AI Coding Plan	umans-ai-coding-plan
IO.NET	io-net
GMI Cloud	gmicloud
Xiaomi Token Plan (China)	xiaomi-token-plan-cn
Zeldoc	zeldoc
Scaleway	scaleway
OVHcloud AI Endpoints	ovhcloud
Friendli	friendli
Tencent TokenHub	tencent-tokenhub
Weights & Biases	wandb
KUAE Cloud Coding Plan	kuae-cloud-coding-plan
GitLab Duo	gitlab
Kilo Gateway	kilo
LucidQuery	lucidquery
Meganova	meganova
Perplexity	perplexity
Amazon Bedrock	amazon-bedrock
Umans AI	umans-ai
Together AI	togetherai
FrogBot	frogbot
OpenRouter	openrouter
Jiekou.AI	jiekou
Nova	nova
Alibaba Token Plan	alibaba-token-plan
Alibaba	alibaba
Databricks	databricks
CrofAI	crof
FastRouter	fastrouter
abliteration.ai	abliteration-ai
Xpersona	xpersona
Azure Cognitive Services	azure-cognitive-services
Baseten	baseten
Atomic Chat	atomic-chat
routing.run	routing-run
AIHubMix	aihubmix
Vertex	google-vertex
NanoGPT	nano-gpt
Moark	moark
Lilac	lilac
Ambient	ambient
Neon	neon
Upstage	upstage
Zhipu AI Coding Plan	zhipuai-coding-plan
Chutes	chutes
MiniMax Token Plan (minimaxi.com)	minimax-cn-coding-plan
DeepSeek	deepseek
Wafer	wafer.ai
MiniMax (minimax.io)	minimax
GitHub Copilot	github-copilot
Clarifai	clarifai
Synthetic	synthetic
iFlow	iflowcn
Xiaomi Token Plan (Singapore)	xiaomi-token-plan-sgp
Claudinio	claudinio

License

MIT