[shekhargulati]

> running things in production/self hosting is more annoying than just paying like 20-200/month

This is an important point. As people rely more on AI/LLM tools, reliability will become even more critical.

In the last two weeks, I've heavily used Claude and DeepSeek Chat. ChatGPT is much more reliable compared to both.

Claude struggles with long-context chats and often shifts to concise responses. DeepSeek often has its "fail whale" moment.

[GTP]

> In the last two weeks, I've heavily used Claude and DeepSeek Chat. ChatGPT is much more reliable compared to both

Which reliability problems did you face? I heard about connection issues due to too much traffic with Deepseek, but those would go away if you self-host the model.

[edude03]

Obviously the reliability problems would go away if you self-host but the "point" is most people rely on external providers because they can't locally run models of similar quality. So what do you do if deepseek cuts you off? For most getting 12? H100s (for the 671b) is essentially impossible.

[creamyhorror]

You use a Deepseek model not hosted by Deepseek, but another provider (e.g. DeepInfra currently). Hopefully a robust provider market will emerge and thrive, even if open models start thinning out.