DeepSeek

Free DeepSeek R1 API

Open reasoning model that rivals o1 — free on several providers.

DeepSeek R1 is an open reasoning model whose chain-of-thought performance approaches OpenAI’s o-series on math and code. Full R1 and its distilled variants are hosted free by several providers, making it the best free way to get genuine reasoning.

12
Free providers
Groq
Recommended
DeepSeekReasoningCodingOpen source

Quickstart — DeepSeek R1 on Groq

Python
from openai import OpenAI

client = OpenAI(
    base_url="https://api.groq.com/openai/v1",
    api_key="YOUR_FREE_KEY"   # from Groq
)

resp = client.chat.completions.create(
    model="llama-3.3-70b-versatile",   # DeepSeek R1 on Groq
    messages=[{"role": "user", "content": "Hello!"}]
)
print(resp.choices[0].message.content)

Providers offering DeepSeek R1 free

Create your own key on any of these. Click through for setup, limits and code.

Groq logo

Groq

groq.com

The fastest free inference on the planet — 500+ tokens/sec.

Free tier No card OpenAI-compatible
Free models
Llama 3.3 70BDeepSeek R1 DistillQwen 3 32B
OpenRouter logo

OpenRouter

openrouter.ai

One key, 300+ models — dozens are completely free.

Free tier No card OpenAI-compatible
Free models
DeepSeek V3 (free)Llama 4 Maverick (free)Gemini 2.0 Flash (free)
GitHub Models logo

GitHub Models

github.com

Free access to GPT-4o, o-series and more — just a GitHub account.

Free tier No card OpenAI-compatible
Free models
GPT-4oGPT-4o minio3-mini
Cloudflare Workers AI logo

Cloudflare Workers AI

cloudflare.com

Serverless inference with a free daily allocation on the edge.

Free tier No card OpenAI-compatible
Free models
Llama 3.3 70BLlama 4 ScoutDeepSeek R1 Distill
SambaNova Cloud logo

SambaNova Cloud

sambanova.ai

High-speed inference for large open models, free to start.

Free tier No card OpenAI-compatible
Free models
Llama 3.3 70BDeepSeek R1Qwen 3 32B
NVIDIA NIM logo

NVIDIA NIM

nvidia.com

Free credits across a huge catalog of hosted open models.

Free credits No card OpenAI-compatible
Free models
Llama 3.3 70BDeepSeek R1Nemotron 70B
Together AI logo

Together AI

together.ai

Free signup credit plus a free Llama endpoint.

Free credits No card OpenAI-compatible
Free models
Llama 3.3 70B Turbo (free)DeepSeek R1Qwen 3
Hugging Face logo

Hugging Face

huggingface.co

Monthly free inference credits across many providers.

Free credits No card OpenAI-compatible
Free models
Llama 3.3 70BDeepSeek V3Qwen 3
Hyperbolic logo

Hyperbolic

hyperbolic.xyz

Low-cost open-model inference with free starter credit.

Free credits No card OpenAI-compatible
Free models
Llama 3.3 70BDeepSeek V3Qwen 3
Fireworks AI logo

Fireworks AI

fireworks.ai

Fast open-model serving with free starter credit.

Free credits No card OpenAI-compatible
Free models
Llama 4 MaverickDeepSeek V3Qwen 3
DeepSeek logo

DeepSeek

deepseek.com

Frontier reasoning at rock-bottom prices (free starter credit).

Low cost No card OpenAI-compatible
Free models
DeepSeek V3.2DeepSeek R1
Scaleway Generative APIs logo

Scaleway Generative APIs

scaleway.com

EU-hosted open models with a free beta allowance.

Free tier OpenAI-compatible
Free models
Llama 3.3 70BMistral SmallQwen 3

Free DeepSeek R1 API — FAQ

Is DeepSeek R1 free to use via API?+

Yes. 12 providers in our directory host DeepSeek R1 with free access, starting with Groq. You create your own free key on the provider’s site.

Which provider is best for a free DeepSeek R1 API key?+

Groq — The fastest free inference on the planet — 500+ tokens/sec. See the full comparison below to match limits and speed to your needs.

How do I call DeepSeek R1 with the OpenAI SDK?+

Set your base URL to your chosen provider’s endpoint (shown on each provider page), use your free key, and pass a DeepSeek R1 model id such as "deepseek-r1".