HACKTICK.AI
FREE AI API DIRECTORY

Free AI APIs,
actually useful.

A quick directory of real free tiers and local runtimes. Find a model, see the catch, then get back to building.

25 cloud providers11 no-account options9 local runtimes

THE DIRECTORY

Pick your free.

How to choose
01G

Google AI Studio

Visit provider
1,500 requests / day17 models
1MGemini prototyping
03NV

NVIDIA NIM

Visit provider
40 requests / minute82 models
1MLarge catalog testing
04OR

OpenRouter

Visit provider
18 models at zero cost18 models
1MOne API, many models
05VG

Vercel AI Gateway

Visit provider
$5 credits / month12 models
1MVercel projects
06AMD

AMD Radeon Cloud

Visit provider
$10 usage / day5 models
1MHigh daily allowance
07RQ
200 requests / day12 models
1MFree model gateway
08CB
$5 one-time credit2 models
65KVery fast tokens
09CF

Cloudflare Workers AI

Visit provider
10,000 neurons / day40 models
262KEdge applications
10M

Mistral AI

Visit provider
Free mode and Labs12 models
256KModels, OCR & code
11CO
1,000 requests / month12 models
256KRAG & reranking
12L7
10 RPM (40 with token)5 models
1MNo-signup testing
13DS
Credits on sign-up3 models
1MReasoning models
14HF

Hugging Face

Visit provider
$0.10 credits / month135 models
131KOpen-model experiments
15MS

ModelScope

Visit provider
2,000 requests / day50 models
1MChinese frontier models
16Z

Z.ai (Zhipu)

Visit provider
GLM Flash models free4 models
200KFree GLM Flash
17SN
20 requests / day7 models
1MFast open models
18OC

Ollama Cloud

Visit provider
Monthly starter credits19 models
1MHosted Ollama
19OV

OVHcloud AI Endpoints

Visit provider
7 models at €07 models
262KEU-hosted API
20AL

Alibaba Model Studio

Visit provider
1M tokens / model / 90 days5 featured models
1MQwen direct from source
21OZ

OpenCode Zen

Visit provider
6 free models6 models
1MCoding-focused gateway
22A21
$10 credits for 3 months2 models
256KLong documents
23SF

SiliconFlow

Visit provider
Per-model caps after KYC3 models
131KChinese open models
24NB

Nebius AI Studio

Visit provider
$1 trial for 30 days1 featured model
128KEU GPU cloud
26OL
No quota. Ever.Unlimited local models
Your hardwareLocal CLI + serverlocalhost:11434/v1
27LM
No quota. Ever.GGUF and MLX models
Your hardwareDesktop local AIlocalhost:1234/v1
28LC
No quota. Ever.GGUF models
Your hardwareLean local servinglocalhost:8080/v1
29JN
No quota. Ever.Local models
Your hardwareOffline desktop chatlocalhost:1337/v1
30G4
No quota. Ever.Local RAG
Your hardwarePrivate document chatLocal RAG
31VL
No quota. Ever.Open-weight models
Your GPUProduction self-hostinglocalhost:8000/v1
32ML
No quota. Ever.MLX models
Apple SiliconM-series Macslocalhost:8080/v1
33LF
No quota. Ever.Single-file model runtime
Your hardwareOne-file local modelslocalhost:8080/v1
34KC
No quota. Ever.GGUF models
Your hardwareCreative writingGGUF

START HERE

Most APIs speak the same language.

For an OpenAI-compatible provider, you usually only change the base URL and API key. Your existing chat-completions code can stay familiar.
client.ts
const client = new OpenAI({
  apiKey: process.env.API_KEY,
  baseURL: "https://api.provider.com/v1"
});
Need privacy? Use a local runtime. Prompts stay on your machine.
Need to ship fast? Start on a cloud free tier, then price production before launch.

Free tiers and limits change. Treat this as a starting point, then verify the provider’s current terms before you build around it.