// free llm api

One endpoint.
Every open-weight LLM.

Free OpenAI-compatible LLM API, with DeepSeek, GLM and NVIDIA Nemotron at no cost. Also Kimi, Qwen, MiniMax and NVIDIA Nemotron, with streaming, function calling and per-token billing in USD.

POST /v1/chat/completions200 OK

$ curl https://api.litepod.io/v1/chat/completions \
  -H "Authorization: Bearer $LITEPOD_API_KEY" \
  -d '{"model": "z-ai/glm-5.3",
       "stream": true,
       "messages": [{"role": "user",
                     "content": "Hi"}]}'

data: {"delta": {"content": "Hello"}}
data: {"delta": {"content": "! How can"}}
data: {"delta": {"content": " I help?"}}
data: [DONE]█

models served: 9; models served
model labs: 6; model labs
max context tokens: 1M; max context tokens
API key for all of them: 1; API key for all of them

// start free

Free to use

Build and test on these models at no cost. Switch to any paid model later by changing one string.

NVIDIAfree

Nemotron Lightning 3.5

nvidia/nemotron-lightning-3.5-30b-a3b

context: 262K tokens
can do: reasoning · tools
limits: 50/day · 10/min

try in playground →

// limits reset daily at 00:00 UTC

// modeldock

ModelDock: test open models on a free GPU

A separate product for trying open models from Hugging Face on a free GPU.

visit lab.litepod.io ↗

// compared

How Litepod compares

Free models and your own models, next to OpenRouter, Together AI and Groq, on facts from their own docs.

Feature	Litepod	OpenRouter	Together AI	Groq
OpenAI-compatible API	Yes	Yes	Yes	Yes
Free models	1 model, 50 requests a day	Free variants, 50 a day, or 1,000 a day after $10 in credits	Lists some $0 models	Free plan, limits per model
Your own Hugging Face model	Free GPU test on ModelDock (separate product)	Not in its docs	Yes, on dedicated endpoints	LoRA adapters

// from each provider's public docs, as of 3 October 2026; check their pages for current terms. Sources: OpenRouter limits, Together AI custom models, Groq rate limits

// openai-compatible

Switch models by changing one string

Point any OpenAI SDK at the Litepod base URL and keep the rest of your code. Pick a model to see the only line that changes.

app.pyDeepSeek

import os
from openai import OpenAI

client = OpenAI(
    base_url="https://api.litepod.io/v1",
    api_key=os.environ["LITEPOD_API_KEY"],
)

reply = client.chat.completions.create(
    model="deepseek/deepseek-v4.1-flash-paid",
    messages=[{"role": "user", "content": "Hello!"}],
)  # only the model string changes

try another model

// long-context inference

Context windows up to 1M tokens

Context length is published per model, and the largest reaches one million tokens. Load entire codebases, contracts or document sets into a single request, and pay only for the input tokens you actually send.

DeepSeek V4.1 Flash: 1M tokens
GLM-5.3: 1M tokens
Kimi K3: 1M tokens
MiniMax M3: 524K tokens
Nemotron 3 Ultra: 262K tokens

GET /v1/models

Production-ready open-weight models

MODEL	LAB	ID
DeepSeek V4.1 Flash	DeepSeek	deepseek/deepseek-v4.1-flash-paid
GLM-5.3	Z.ai	z-ai/glm-5.3
GLM 5.3 Flash	Z.ai	z-ai/glm-5.3-flash
MiniMax M3	MiniMax	minimax/minimax-m3
Nemotron 3 Ultra	NVIDIA	nvidia/nemotron-3-ultra
gpt-oss-120b	OpenAI	openai/gpt-oss-120b
Nemotron Lightning 3.5	NVIDIA	nvidia/nemotron-lightning-3.5-30b-a3b
Kimi K3	Moonshot AI	moonshotai/kimi-k3
Qwen 3.8 Max	Alibaba	qwen/qwen3.8-max
Kimi K3.1 Pro	Moonshot AI	moonshotai/kimi-k3.1-pro
MiniMax M3.1 Flash	MiniMax	minimax/minimax-m3.1-flash

Drop-in compatible

Point any OpenAI SDK at api.litepod.io/v1. Streaming, vision and function calling work as-is.

Long context

Context length is published per model, and the largest reaches one million tokens.

Metered in USD

Pay per input and output token. Usage and spend per key, per model, in your dashboard.

One endpoint.Every open-weight LLM.