Skip to content

Free providers

NOTE

Providers in this list offer access to models absolutely for free, but may have certain restrictions (limits).

NVIDIA NIM

Requirements Limits 🔗Models Reliability Testing
✉️
🔑
📱		 RPM - 40 🌐
• DeepSeek v4 flash/pro
• kimi-k2
• glm-4.7/5.1		 ⭐⭐⭐ ✅
🔗 Registration 🔗 API key 🔗 Base URL
🌐 🌐 https://integrate.api.nvidia.com/v1

Disadvantages

  • Problematic phone verification (SMS can be delayed)
  • Models are unstable during peak hours (US daytime)
  • Requires an Nvidia account (no graphics card needed)

LLM7

Requirements Limits 🔗Models Reliability Testing
📝❗ • RPM - 20
• RPH - 100		 🌐
• GLM-4.6V-Flash
• codestral-latest		 ⭐⭐ ✅
🔗 Registration 🔗 API key 🔗 Base URL
🌐 🌐 https://api.llm7.io/v1

Disadvantages

  • Few models available.
  • Unstable list of models (they may disappear).

Google AI Studio

Requirements Limits 🔗Models Reliability Testing
✉️
🔑		 Individual limits for each model
🌐		 🌐
• Gemini 3 Flash/Flash Lite
• Gemma 4 31B/26B		 ⭐⭐⭐⭐⭐ ✅
🔗 Registration 🔗 API key 🔗 Base URL
🌐 🌐 https://generativelanguage.googleapis.com/v1beta/openai/

Disadvantages

  • The best models are not available

Airforce

Requirements Limits 🔗Models Reliability Testing
🔑 • RPM - 1
• RPD - 1000		 🌐
• 130+		 ⭐⭐⭐⭐ ✅
🔗 Registration 🔗 API key 🔗 Base URL
🌐 🌐 https://api.airforce/v1

Features

  • Sign up using my link to get 1 day of Premium

Disadvantages

  • Very low RPM.

Logfare

Requirements Limits 🔗Models Reliability Testing
🔑 - 🌐
• DeepSeek V4 Flash
• Gemini 3 Flash
• Kimi K2.6		 - ⏳
🔗 Registration 🔗 API key 🔗 Base URL
🌐 🌐 https://logfare.ai/v1

Features

  • There are no limits.

Disadvantages

  • They collect your data such as IP, promt and more. They then transfer it to third parties (anonymously). Keep this in mind if this is important to you.
  • NSFW is allowed, but only if it's not prohibited by law. Otherwise, ban

Zydit

Requirements Limits 🔗Models Reliability Testing
✉️
🔑		 • RPM - 10 180+
🌐		 ⭐⭐⭐⭐ ✅
🔗 Registration 🔗 API key 🔗 Base URL
🌐 🌐 https://api.zydit.in/v1

Features

  • Not exactly a provider and not exactly a proxy; Zydit doesn't host models itself, but uses only official providers.
  • Very large list of models
  • Currently in testing, completely free

Disadvantages

  • Very low RPM.

Groq

Requirements 🔗Limits 🔗Models Reliability Testing
✉️
🔑		 🌐 🌐
• gpt-oss-120b
• kimi-k2-instruct-0905		 ⭐⭐⭐⭐⭐ ✅
🔗 Registration 🔗 API key 🔗 Base URL
🌐 🌐 https://api.groq.com/openai/v1

Disadvantages

  • Few models available

Cerebras

Requirements 🔗Limits 🔗Models Reliability Testing
🔑
✉️		 Individual limits for each model
🌐		 🌐
• gpt-oss-120b
• qwen-3-235b-a22b-thinking
• glm-4.7		 ⭐⭐⭐⭐ ✅
🔗 Registration 🔗 API key 🔗 Base URL
🌐 🌐 https://api.cerebras.ai/v1

Disadvantages

  • Some models have a smaller context than the official provider.
  • Besides request limits, there are token limits.

Mistral

Requirements Limits 🔗Models Reliability Testing
✉️
🔑
📱		 🌐
Individual limits for each model		 🌐
• Mistral Large 3
• Devstral 2
• Mistral Medium 3.1		 ⭐⭐⭐⭐⭐ ✅
🔗 Registration 🔗 API key 🔗 Base URL
🌐 🌐 https://api.mistral.ai/v1/

Disadvantages

  • To get the free tier, you must consent to the use of your data.

BlazeAI

Requirements Limits 🔗Models Reliability Testing
📝❗ • RPD - 1000 🌐
• DeepSeek V4 Flash
• Kimi K2.5/2.6
• MiniMax M2.7
• claude-opus-4.5
• qwen3.6-max/plus		 ⭐⭐⭐ ✅
🔗 Registration 🔗 API key 🔗 Base URL
🌐 🌐 https://blazeai.boxu.dev/api/

Disadvantages

  • There are issues with the qwen 3.6 model family. It detects the tools but cannot use them. (If you're only into roleplay, this won't be a problem for you.)

More here free-llm-api-resources - just a list.