Since many people, myself included, can't afford paid LLM coding plans, I often look for free community-run LLM API services. I'm sharing a collection here today, including services for models such as ChatGPT, Claude, Gemini, Grok, GLM, Mimo, and Kimi.

Because these services are community-run, their availability is not guaranteed. I still get plenty of use out of them from time to time, so I'm sharing them here. I'll update this list periodically. I personally verified that the sites below were working at the time of writing, but that doesn't mean they'll remain available. If you find that an offer has expired or a service is no longer working, please let me know in the comments and I'll cross it off the list. The last two have been available for a long time. I don't know which wealthy benefactor provides them, but I've been using them myself.

JustDoWork

The JustDoWork community service only offers the claude-opus-4-8 model. Sign up to receive $100 in credits. Register now at https://api.justwoker.icu/register. As of 09.27, my test showed that it was working:

Credits offered by JustDoWork
Credits offered by JustDoWork
JustDoWork Opus4.8
JustDoWork Opus4.8

Vyce AI

Vyce AI isn't a community service; it's a gateway for accessing LLMs, and you can pay for it if needed. New users currently receive $40 in credits, and daily check-ins earn another $10. Sign up at https://vyceai.com/signup. Note that Discord authorization is currently required to access all models.

Vyce AI
Credits offered by Vyce AI

Let's test the deepseek-v4.1 model. It was working as of 09-25.

vecy ai deepseek
vecy ai deepseek

CodeCraft

CodeCraft provides 1 million tokens each month. That's not a huge amount, but its models are genuinely good: GPT-5.6 Luna, the full Claude lineup, and a wide range of high-end Chinese models. Updated on 2026-09-25. If you're interested, sign up at https://codecraftapi.com. A friendly reminder: two accounts give you 2 million tokens!

CodeCraft
CodeCraft
Models offered by CodeCraft
Models offered by CodeCraft

Let's configure it using an official API key. I'll test it with claude-sonnet-5. I couldn't get it to work with Claude itself, but it works with opencode.

Claude Opus 4.8
Claude Opus 4.8

evomap

New users get 1,500 points. I couldn't find an expiration date. It supports the new Kimi K3 model, and you need to download the client to claim the points. Visit evomap.

Free benefits include:

  • Basic EvoX usage
  • Use 20 EvoMap assets in EvoX
evomap
evomap

Mimo

This AI coding terminal is an adaptation of OpenCode developed by Xiaomi. MiMo-V2.6-flash is currently free to use. It's an official, reliable option; just install the mimo cli.

Mimo
Mimo

anyrouter

I don't know which mysterious benefactor is behind this site. It supports official Claude models, including the latest Claude Opus 4.8. I used it frequently at first, and I've accumulated $3,000 in check-in credits so far.

Unfortunately, there are too many users, so the service is often unavailable. I've heard it may only work late at night, but I go to bed early, and GPT is good enough for me. Sign up and give it a try if you're interested. Visit: https://anyrouter.top

If you keep getting an API Error, try setting this environment variable: ANTHROPIC_CUSTOM_HEADERS: x-stainless-os: Windows

{
"env": {
"ANTHROPIC_CUSTOM_HEADERS": "x-stainless-os: Windows"
}
}

agentrouter

agentrouter is another long-running community service from around the same time as anyrouter. It currently offers Claude, gpt5.5, glm5.2, and other powerful models. Visit https://agentrouter.org/login.

agentrouter
agentrouter

## Your Community Service

This one is fairly stable and requires an L account to log in. It supports GPT but doesn't offer daily check-ins. You need to top up with LDC, which is effectively free; the only downside is that there are no check-in credits. Visit: https://muyuan.do

Summary

Community-run LLM API services are made available for free by their operators, so they aren't always reliable. They often go offline when upstream accounts are restricted or their quotas are used up. Still, you can't complain about a free service. If one stops working, leave a comment and I'll remove it. If you know of other free community-run LLM API services, share them in the comments and I'll add them to the article.