51–60 of 157 posts
Technical report is available here https://www.amazon.science/publications/the-amazon-nova-fami...
TL;DR comparison of models vs frontier models on public benchmarks here https://imgur.com/a/CKMIhmm
The one that really stands out is GroundUI-1K, where it beats the competition by 46%.
Nova Pro looks like it could be a SOTA-comparable model at a lower price point.
It's really amusing how bad Amazon is at writing and designing UI. For a company of their size and scope it's practically unforgivable. But they always get away with it.
A rough idea of the price differences... Per 1k tokens Input | Output Amazon Nova Micro: $0.000035 | $0.00014 Amazon Nova Lite: $0.00006 | $0.00024 Amazon Nova Pro: $0.0008 | $0.0032 Claude 3.5 Sonnet: $0.003 | $0.015 Claude 3.5 Haiku: $0.0008 | $0.0004 Claude 3 Opus: $0.015 | $0.075 Source: AWS Bedrock Pricing https://aws.amazon.com/bedrock/pricing/
Has anyone found TPM/RPM limits on Nova? Either they aren't limited, or the quotas haven't been published yet: https://docs.aws.amazon.com/general/latest/gr/bedrock.html#l...
It's really amusing how bad Amazon is at writing and designing UI. For a company of their size and scope it's practically unforgivable. But they always get away with it.
For most of AWS offerings, it literally doesn't matter and logging in to AWS Console is a break glass thing.
Case in point: this very article. It uses boto3 to interface with AWS.
It's really amusing how bad Amazon is at writing and designing UI. For a company of their size and scope it's practically unforgivable. But they always get away with it.
You say they "get away with it," but it makes more sense to conclude that UI design has a lot lower ROI than we assume it does as users.
Does this mean they trained multiple copies of the models?
This is blowing my mind. gemini-1.5-flash accidentally knows how to transcribe amazingly well but it is -very- hard to figure out how to use it well and now Amazon comes out with a gemini flash like model and it explicitly ignores audio. It is so clear that multi-modal audio would be easy for these models but it is like they are purposefully holding back releasing it/supporting it. This has to be a strategic decision to not attach audio. Probably because the margins on ASR are too high to strip with a cheap LLM. I can only hope Meta will drop a mult-modal audio model to force this soon.
A rough idea of the price differences... Per 1k tokens Input | Output Amazon Nova Micro: $0.000035 | $0.00014 Amazon Nova Lite: $0.00006 | $0.00024 Amazon Nova Pro: $0.0008 | $0.0032 Claude 3.5 Sonnet: $0.003 | $0.015 Claude 3.5 Haiku: $0.0008 | $0.0004 Claude 3 Opus: $0.015 | $0.075 Source: AWS Bedrock Pricing https://aws.amazon.com/bedrock/pricing/