Free LLM options (that are compatible with CHub frontend)
There are already several resources out there listing general free API options, so this is specifically focused on those that can be used on Chub.
27/07/25 update:
Chub now supports proxy URLs! That means you can plug any OpenAI-compatible endpoint into the Reverse Proxy URL field to use external models from that API. Not all APIs work with Chub, some will give errors because of a fundamental incompatibilty with the way Chub works. I haven't tested every single known free provider, obviously, so please tell me if you find a working one so I can add it.
Current free APIs confirmed not to work with Chub:
- Mistral
- Nvidia
Chub Free Model
Daily messages: unlimited
Censorship: none
Free, unlimited, and uncensored. The downside is that it's used by the devs to test potential new models for the paid tiers, so quality can change at any time. Because of that, there's not much point reviewing the quality since it's up to what the devs are testing.
As of writing (July 2025), it's temporarily frozen at Mistral 7B since the devs aren't testing anything. For comparison, Deepseek V3 is a whopping 671B, so adjust your expectations accordingly. You can look at my old Chub models review for a quick rundown of it.
Openrouter
Daily messages: 50/day (free), 1000/day (10USD topup)
Censorship: varies, but Deepseek and most roleplay models are uncensored
OR gives 50 total free messages a day, shared across all their free models. Topping up 10USD into the account increases the limit to 1000 total free messages/day, and the credit lasts for a year.
Important: you DO NOT need to keep a 10USD standing balance in the account. As long as you've topped up a total of 10USD since the account was made, the 1000/day unlock is permanent. Feel free to spend the credit on trying out paid models, it won't reduce your free message limit.
There are several providers which offer Deepseek, both V3 and R1, for free through OR. Kimi K2 is another popular corpo-sized open model.
Chutes (no longer available)
Daily messages: 200/day (5USD topup), rerolls cost 1/10 of a message
update: Chutes has confirmed via their Discord that they will be throttling free requests from OR to prioritize direct users instead. While $5 for 200/day is still objectively less value than $10 for 1000/day, which to choose depends on your own model usage. If you intend to use a model for which Chutes is the only provider, the $5 will ensure you don't hit rate limit errors when using it during peak timings.
update 2: Chutes has removed their own free tier entirely as of August 2025, and transitioned entirely to a monthly subscription model. The only way to use Chutes for free now is having an account that did the $5 topup before it was removed, or through Openrouter.
Chub now supports proxy URLs natively so you can theoretically use Chutes directly now, although I don't know if it works with Chub since I don't have a $5 Chutes account. Someone test and tell me, I guess.
Honestly, don't use Chutes. $5 for 200/day from only 1 provider, which has a history of occasionally going down, is objectively worse than $10 for 1000/day from multiple providers.
Gemini
Daily messages: varies by model
Censorship: medium
Gemini is listed as Palm2 in the Chub secrets menu, since that's what it used to be called before it was Gemini.
Gemini Pro is the big smart model, Flash is the faster but dumber model, and Flash Lite is the really small but much dumber one. Check the link for individual limits—the bigger and smarter models naturally have fewer messages/day than the smaller ones. Personally, don't bother with using models that aren't from the 2.5 line; the improvement of the 2.5 models over the earlier ones are significant enough that the earlier ones are frankly just bad in comparison.
Being a corpo model, Gemini is obviously censored, although the filters are configurable and Chub sets them all to minimal. Still, there are certain topics like underage and non-con that are hard filtered by Gemini no matter what. It's possible to jailbreak it into allowing those through, but I won't talk about exact methods here since those tend to change with every model update.
Gemini (via Openrouter)
Censorship: HIGH
Don't use Gemini through OR. Reason being that while most RP frontends set Gemini's filters to as low as possible, OR doesn't. You have no control of Gemini's filter level through OR, which sets them all to medium.
Groq (via proxy URL)
Daily messages: varies by model
Censorship: varies by model, mostly uncensored
It's Groq with a Q, not Grok with a K. Groq is a provider which hosts models on their hardware. Don't get it mixed up with Xitter's model which is most certainly not free. You'll need to use this via proxy URL, but Groq provides limited free access to a number of foundational open models.
Pawan
Daily messages: check with Pawan directly
Censorship: unknown
All I will say about Pawan models is that they exist as an option. Chub mods and community helpers have chosen not to provide any support for Pawan models due to past antagonistic behavior on their end. Any questions about their models and usage limits should be directed to their Discord server, which you can easily Google.
If I missed anything, or there are any new free providers you want added, I can be found on the Chub discord as @kirandra or at kiranyandra@proton.me