“I first tried to access dsv4 models from openrouter pinpointing to selected providers only - lots of issues, cache not working always and slow responses. Then switched to using directly their api - all problems disappeared.”
“If you have an old computer around you can download openwebui without docker and connect it to openrouter and use tailscale to use it on your laptopo while out side”
“I'm currently using the OpenRouter Web Chat and there are a few features I find extremely useful, but the lack of persistence makes the web version unusable for me long-term. Chats are gone after a reload, there's no history or project structure, nothing to look back on.”
“I initially used OpenRouter with DeepInfra and Together AI. The results were much too noisy to draw reliable conclusions from, and markedly different behavior through AWS Bedrock was the last straw.”
“Openrouter inference is cheap, it's extremely capable for nearly everything I do, and works fast in my experience. When I can someday get this kind of performance/speed on my local hardware I'll be very, very happy.”
“Another thing to add that others haven't touched on is cache hit rate. For me, it's not been great for DS V4 Flash, around 80% at best as reported by OpenRouter. On long sessions with a ton of turns and tool calls it ends up being more expensive that GLM-5.3-Flash because of all the expensive cache misses at 200k+ depth.”
“Openrouter currently lists non-peak In / Out Price of $0.05 / $0.16 per 1M for this model (via openinference), while official Deepseek pricing is $0.22/$0.66 which is a good 4x higher. I guess I will note that cached read is $0.007 from DS direct while openinference gives $0.013 for that.”
“it is very intuitive and openrouter is linked to it too if you wanna switch the free gemini models from openrouter free model. wi might link another site as well to it who have all models that have dissapeared from our reach too.”
“I'll stop paying for Claude Pro this week, moving on to OpenCode + OpenRouter models or Reasonix + DeepSeek official API credits, which are cheap even with the price increase. They released a new DeepSeek model today BTW.”
“Use GPT 5.6 Luna with your Plus subscription and GLM 5.3 Flash via OpenRouter. Use the Cline extension inside VSCode and Hermes if doing anything agentic.”
“Idk, you’d have to use it an insane amount in my opinion for a subscription to be worth it, plus then you’re locked in to that set of models vs openrouter you just choose whatever you want.”
“So yeah, for now I'm happy with Openrouter, because I don't have to spend 12 USD to do my roleplay each month. However, if Deepseek flash raises it's prices considerably or Minimax M3 or Nemotron Ultra are discontinued I may jump ship and bite the bullet of Nanogpt sub.”
“But watching active money tick down when playing on OpenRouter diminishes my playing experience more than just knowing I have a $12 sub with a limit I'm realistically not going to hit so I can just disregard it.”
“But you can just throw 10$ or so on openrouter and see how far you get with it, it's really not that big of a deal. I think the new gemini is def worth it every once in a while.”
“But now, I’m seeing all the hype around Gemini 3.8 Flash, and I noticed it’s on the list of models available on OpenRouter. I’m looking for advice and recommendations from OpenRouter users—what’s it like?”
“Definitely give all the settings in OpenRouter a once-over, especially the Tools page. In my case, their third-party MCP tools like WebSearch and WebFetch ate up my tokens pretty quick.”
“Openrouter is not so open as well, right? They take the middle man cut by enforcing their own limit restrictions and 429s on top of the model provided limit restrictions and what not.Now why would I pay a middle man anything if I can get the same hosted model without their own whimsical service limits?”