DEV Community

Cover image for I Tested 100+ Providers to Find Free and Replenishable Text-generation LLM APIs
spaceyyy
spaceyyy

Posted on

I Tested 100+ Providers to Find Free and Replenishable Text-generation LLM APIs

I searched around hundreds of LLM Providers across the web and GitHub, To find LLM APIs that have a free tier and should automatically refill itself on a cycle, whether it could be daily, weekly or monthly.

This all started back where I wish to have a LLM API that is needed for my long-term projects. There are so many options I can choose from, but sometimes I just needed a lot of "fallbacks" to support my project.
But to prevent it from failing, the LLM APIs are needed to replenish itself, which considering I found a few of them and they are extremely common, It felt like it was not enough.

Having to add some providers which would have a one-time credit or a free trial would really not fit it, considering it would be drained in the long run, and never be able to use it again.

So, I've gone throughout the web, sailing across websites that promise Free LLM APIs to follow these two main goals:

  • To be Free without billing info.
  • To be Replenishable without one-time or trial credits.

I also have a third goal if second condition fails: If it's not Replenishable, it must have at least one free model available in their catalog that works without wasting a single credit dime (>=$0.0000001), and without asking to require credit top-ups.

I went through several popular AI agents with their provider documentation showing a list of providers, so that my goal is to sign myself up as a normal, unpaid user whose requirement is to know whether it refreshes its free quota or not. Not only that, I did explore a lot of GitHub Repositories finding providers that curators add onto their lists.

After I found many LLM providers that "promises" those two rules, it is not the end yet. So, I stress-tested their model endpoints from each provider according to their rate limits.

Surprisingly, the number of LLM Providers has dropped due to not having a single model found to be 200 OK but requiring payment. However, not all providers are dropped when some providers have one or more models recieving 200 OK from just a single prompt, without spending tokens.

To store it as helpful information, I made a repository that showcases all verified providers and its models under a single README. The number of providers that I found across the internet seemed to increase slow and steady overtime due to providers that are seen as obscure, but follows the main two golden rules that keeps the list on-topic as it was.

Hence, it is named "Free BYOK Models", where the theme suggests for "Free AI models from their BYOK-compatible LLM Providers that are free and replenishable."
This is solely for Text-generation endpoints, but it can support Vision-compatible models too. It is useful for trying to integrate API keys using these providers into their AI assistants, Coding agents, etc.

The following table shows available LLM providers along with their Free Tier Quota, at the time of writing:

LLM Provider Free Tier Quota
AION Labs 15 RPM / 20,000 TPD
Agnes AI 20 RPM / 1,000 RPD
AnyAPI AI 100,000 tokens/day / No Credit Card
Auriko 500 RPM (BYOK) / 1,000 RPM (Platform) / 1,000,000 tokens/month (BYOK) / Has Permanently Free models
BazaarLink 10 RPM / 50 RPD / Free Models only
Cloudflare Workers AI 150 to 1,500 RPM / 100,000 RPD / 13,000 TPD
Cohere AI 20 RPM / 1,000 API calls per month
ElectronHub 5 RPM / $0.25 Weekly Credits
EvolveX 5 RPM / No Credit Card
FastRouter 10 RPD per model / No Billing Credits Required
Free.ai 10 RPM / 30,000 TPD / 1,000 Requests per month / Currently available self-hosted models only
FreeInference $20 CPD / 2 Max Concurrent Requests
Google Gemini 5-20 RPM / 20-500 RPD / 1M TPM / Uncapped TPD
Gonka Broker 6 RPM / ~1M tokens per month
Groq API 30 RPM / 14,400 RPD / 18,000 TPM
HelixMind 3 RPM / 50 RPD
Hugging Face Inference API $0.10/month credits (~650K tokens)
Intern AI 30 RPM / 300,000 TPM / 90,000,000 Tokens per month (3,000,000 TPD)
Kilo Gateway 5 RPM / 200 RPD
LLM.Kiwi 40 RPH / No Credit Card
LLM7.IO 40 RPM / 2,400 RPD / 128,000 Characters per Request / 1,000,000 TPD
LiteRouter Unlimited Requests (for some Free models) / 1 concurrent request / 7s Cooldown
MegaNova AI 60 RPM / 550 RPD / 200,000 TPM
Mistral AI ~2–30 RPM / 50,000 TPM shared pool
Mixlayer 20 RPM / Can be rate-limited (daily usage)
Naga AI 10 RPM / 100 RPD
NVIDIA NIM 40 RPM / Uncapped TPD
Odirouter 5 RPM / 50 RPD / Free Models Only / 2 Parallel Multimodal Queries
Ollama Cloud 1 Instance / 5-Hour Session Usage / 7-day Weekly Usage
OpenCode Zen 30 RPM / 500 RPD / 1,000,000 TPD / Daily Limits
OpenRouter 20 RPM / 50 RPD
Orcarouter Unspecified rate limits / Free models only
Poixe AI 10,000 RPD / 10,000,000 TPD
Pooled AI 1M TPD / Minimax models only
Poolside 20 RPM / 200 RPD / 150,000 TPM / 1,000,000 TPD
Requesty 200 RPD / Free Models only
Routeway AI 5 RPM / 200 RPD / 300,000 TPD
SEA-LION 10 RPM
TokenReply 3 RPM / Free Models Only
Void AI 100 RPM / 125,000 Daily Credits
xKiro AI 5M TPD / Free models only
Yolo-Auto 15 RPD
Z.AI (Zhipu AI) 1 Concurrent Request / Uncapped TPD
Zydit AI Unlimited Requests / 10 RPM / Free models only (For v3 endpoints)
Zylo API 10 RPM / 7,200 RPD / 200,000 TPD

The repository below is a curation list that contains model lists of verified LLM providers along with their latencies, Context Windows, free tier quotas, and Base URLs.
At the time of writing this post, there are 45 LLM Providers that resets their quota in a time cycle.

Contributing

The repository is updated regularly to make it up-to-date with its latest model lists. If you find a provider that is free and replenishable, and has no gated services, feel free to comment, or submit a PR since it updates fast.

Star the Repository if you find this useful!

Top comments (0)