20 comments

[ 4.4 ms ] story [ 40.5 ms ] thread
Absolute gem. And after recent tweak this table is fast af even it is huge with a lot of rows.
This seems really interesting, thanks maxloh for sharing.

I have a quick question but https://aihubmix.com/model/coding-glm-5.1-free seems to be free in the chances of "coding-glm-5.1-free is the open and free version of coding-glm-5.1. To ensure stable service performance, usage limits are in place: up to 5 requests per minute, 500 requests per day, and a daily token allowance of 1 million."

Is there any catch aside from that for this aihubmix? I use opencode-zen from free version mostly if I want agents but this seems interesting to me as well and I think that it mught be able to get integrated into opencode itself as well given this repo is from opencode (well anomalyco)

A quick question but is there any tangible benefit of using these AIhubmix or others over something like opencode-zen itself that I may be missing?

Really useful database, though the website could really use a filtering feature rather than just sorting.
Interesting approach. The unified pricing table is helpful, but I'd love to see latency benchmarks across providers – that's often the hidden cost beyond price/token.
Its interesting and pretty useful if it can be kept updated in long run. Also some way to capturing the changes in pricing if any given that one of the arguments is that over the time the prices should come down.

nit suggestion: It took a while to realize that I have to scroll to right to see more details. Most users are in habit of scrolling down or click on some button to see more details.

Definitely needs filtering for all the data, so you can block out "closed" models, and even models that are not LLMs.
This is missing data like when particular model was nerfed or how often provider routes to cheaper less capable model (variants of so called adaptive reasoning).

Cost per token says nothing. For instance if model goes dumb half way the task and you have to start again. If model does that all the time, then the cost is substantially higher than headline figure.

Probably such a service should constantly run various types of tasks on such models and gauge quality of output (though still provider can detect it and pin their best model to skew the results).

Publishing a table without filter features should be illegal
Useful concept but missing some obvious features like filtering on availability of tool calling or support for different modalities, etc.
What would be helpful is also harness + models. I realized I can get Pi Dev and DeepSeek Pro to do the same work as Claude Code + Opus 4.7.

I was skeptical at first, but I tried using that and then asked Claude if the changes were okay, and they were, and the code works fine.

This needs filter by “current”. Most of models are superseded by newer versions that cost same or less, and there’s no reason to have old ones in the list when most of users want to evaluate the current market.
cool, but the graveyard of abandoned "comprehensive AI model databases" is already pretty crowd, I think
It's particularly pernicious that different models have different structured information encodings.

In general these systems feel so so so near to each other. But endless little foibles & differences. Its ridiculously difficult.