10ms. Unlimited for $20/month. Now multimodal.
Today, @OctenAI is launching:
1. Octen Pulse: Open-source search at the speed of a single token.
2. Unlimited Search: $20/month. No more counting API calls.
3. Multimodal Search: Text and images. Platform.
Humans search with Google. Machines search with Octen.
hm.. there are ~3,700 emojis, and unicode already provides keywords for each. (Even if we use all of them). “should do well” ain’t the point, “should we use it” is the point. A forklift can also lift a coffee mug :)
maybe thinking like “if I didn’t work for Qdrant how would I solve it?” is a better question to give you a few options.
@antoine_chaffin Interesting, what are some real-world corner cases you have encountered / working on where frontier throughput is attractive and sub-sec latency is less of a or not a concern.
anything particular standouts ?
To squeeze more performance authors can
do a simple model soup (averaging weights from different training checkpoints or hyperparameter runs of the same architecture) works well in embedders and retrievers , science is strong here - multiple cases to cite.
In mergekit I believe it is linear merge.
Other fancy tricks are needed only for cases like: blend specialists trained on different tasks and you want to easily borrow that strength.