Does work with openrouter.ai?

[almagest21]

Thanks for this brilliant project.
It was very difficult to determine where the problem was coming from, but I would like to report it.

I experienced this problem when used with OpenRouter.ai's OpenAI compatible endpoint (https://openrouter.ai/api/v1/chat/completions).
As shown below, I receive choppy reception with missing characters.
If the streaming option is turned off, it appears normal.
(When adding Custom Models, the [Enable streaming for real-time response generation] option)

Could you please confirm this problem?

Model: gpt-4o-mini
Max Token: 4000
Temperature: 1
Top-p: 1
Presence Penalty: 0
Frequency Penalty: 0
Image Detail: auto

System:
You are ChatGPT, a large language model trained by OpenAI.
Carefully heed the user's instructions.
Respond using Markdown.

User:
Please describe your model

Assistant:
Sure! amGPT a language model byAI Here some aspects my:

Architecture-Type: on GPTGener Pre Transformer architecture-Layers: consist multiple of, help understanding context and relationships within the.
Training-Data: have trained dataset includes wide of,,, more-Objective: training is predict next in sentence enabling to coherent context relevant.

Cap

    ** Language**: I comprehend generate in way resembles conversation-Vers** I assist various including questions, creative, more-Customization: can the or tone of response and can to contexts###ations-Knowledgeoff: training goes until 3 so may not be of or that after date-No-time** I not from or user, privacy confidentiality-Occ Inacies: I for, can provide or.

am to helpful informative, and, a of and. there's specific would to about functionalities feel to!

[animalnots]

I've tested azure openai gpt-4o-mini and it works fine using streaming mode (using config you provided). I don't get choppy responses. Perhaps worth contacting openrouter to address the issue.

[almagest21]

Appreciate your quick response.
Yes, I also believe there is no problem with the official OpenAI API and the Azure OpenAI API. Perhaps there are very slight differences between OpenRouter.ai and these services.

I see that there is a Discord community for OpenRouter, so I will try to post this. Thanks.
If you have time, could you please also try OpenRouter's API.

[animalnots]

Do they have free plan or do I need to pay to get a key?

[almagest21]

I believe there shouldn’t be any payment required unless you’re using a paid model. There are plenty of free models compatible with OpenAI.
Once you obtain an OpenRouter account and API key, you just need to specify a custom endpoint and add a custom model in BetterChatGPT-Plus settings view as you know.

Example:
Set the following as the model ID:
meta-llama/llama-4-maverick:free