Sold by: Liquid AI
Deployed on AWS
Free Trial
Liquid LFM 40B is designed to handle complex tasks, offering an optimal balance between size and output quality.
Overview
Liquid LFM 40B strikes a unique balance between model size and output quality. With 12 billion activated parameters, it delivers performance comparable to larger models while its MoE architecture ensures higher throughput and cost-efficient deployment on accessible hardware.
The model excels in areas such as general and expert knowledge, mathematics, logical reasoning, and long-context tasks. Its primary language is English, but it also demonstrates multilingual capabilities in Spanish, French, German, Chinese, Arabic, Japanese, and Korean.
However, certain limitations exist. The model may struggle with precise numerical calculations, time-sensitive information, or unconventional tasks like counting specific letters in a word. Human preference optimization techniques have yet to be fully implemented, leaving room for further enhancement.
This product is specifically optimized for peak performance on L40S GPUs.
Highlights
- **Innovative Model Architecture**: Liquid AI's Foundation Models utilize a unique architecture that combines liquid neural networks and non-transformer designs, allowing these models to be efficient in memory usage and capable of handling sequential data, such as text, video, and real-time signals. This setup optimizes performance while minimizing computational demands.
- **Enhanced Adaptability and Real-Time Learning**: Unlike conventional models, LFMs can adapt their internal processes based on new inputs in real time, making them highly responsive.
- **Efficiency in Long-Context Processing**: Liquid AI's models can efficiently process extended input sequences without the steep memory and processing requirements typical of transformer-based models, supporting applications like document summarization and complex chatbot interactions with minimal hardware demands. With LFMs, it’s possible to fit up to 1 million tokens-worth of data and map it onto 16 gigabytes of memory. **
Details
Sold by
Delivery method
Latest version
Deployed on AWS
New
Introducing multi-product solutions
You can now purchase comprehensive solutions tailored to use cases and industries.
Features and programs
Financing for AWS Marketplace purchases
AWS Marketplace now accepts line of credit payments through the PNC Vendor Finance program. This program is available to select AWS customers in the US, excluding NV, NC, ND, TN, & VT.
Pricing
Free trial
Try this product free for 7 days according to the free trial terms set by the vendor.
Pricing is based on actual usage, with charges varying according to how much you consume. Subscriptions have no end date and may be canceled any time.
Additional AWS infrastructure costs may apply. Use the AWS Pricing Calculator to estimate your infrastructure costs.
Dimension | Description | Cost/host/hour |
|---|---|---|
ml.g4dn.12xlarge Inference (Batch) Recommended | Model inference on the ml.g4dn.12xlarge instance type, batch mode | $21.8842 |
ml.g6e.12xlarge Inference (Real-Time) Recommended | Model inference on the ml.g6e.12xlarge instance type, real-time mode | $21.8842 |
Dimensions summary
You pay by the hour based on the compute instance you run the model on, with no upfront commitment. Two options are available, each tied to a different instance type and inference style. One runs batch inference on the ml.g4dn.12xlarge instance, which processes grouped requests. The other runs real-time inference on the ml.g6e.12xlarge instance, which handles live requests as they arrive. You choose the instance and mode that fit your workload. Costs scale with the number of host hours you use on the selected instance.
Top-of-mind questions for buyers
You pay for each hour the chosen GPU instance runs the model. One host hour equals one hour of one running instance. The ml.g4dn.12xlarge and ml.g6e.12xlarge are AWS GPU compute types. Meter time starts when the instance runs, regardless of how many requests it serves.
The ml.g4dn.12xlarge batch option processes grouped requests together, which suits high-volume jobs run in scheduled runs. The ml.g6e.12xlarge real-time option handles live requests as they arrive, which suits interactive or low-latency use. Both meter by running host hours, so cost tracks how long each instance stays active.
Charges accrue per running host hour on the selected instance. If you stop the instance, software host-hour charges stop. An idle but running instance still meters time, since billing tracks running hours rather than request volume. Underlying AWS storage or other resource fees may still apply separately.
docs.liquid.ai
Helpful?
Vendor refund policy
We don’t offer refunds, but we’re happy to assist! Contact us anytime at support+aws@liquid.ai .
Legal
Vendor terms and conditions
Upon subscribing to this product, you must acknowledge and agree to the terms and conditions outlined in the vendor's End User License Agreement (EULA) .
Content disclaimer
Vendors are responsible for their product descriptions and other product content. AWS does not warrant that vendors' product descriptions or other product content are accurate, complete, reliable, current, or error-free.
Delivery details
Amazon SageMaker model
An Amazon SageMaker model package is a pre-trained machine learning model ready to use without additional training. Use the model package to create a model on Amazon SageMaker for real-time inference or batch processing. Amazon SageMaker is a fully managed platform for building, training, and deploying machine learning models at scale.
Deploy the model on Amazon SageMaker AI using the following options:
Version release notes
Bedrock release
Additional details
Inputs
- Summary
The model leverages OpenAI's chat format as detailed in OpenAI API documentation , with the following key specifics:
- Supports text-only interactions.
- Input MIME type
- application/json
https://github.com/Liquid4All/aws-samples/blob/main/notebooks/sagemaker/data/input/real-time/simple.json
https://github.com/Liquid4All/aws-samples/blob/main/notebooks/sagemaker/data/input/real-time/simple.json
Input data descriptions
The following table describes supported input data fields for real-time inference and batch transform.
Field name | Description | Constraints | Required |
|---|---|---|---|
Request body | https://platform.openai.com/docs/api-reference/chat | Type: FreeText | Yes |
Resources
Vendor resources
Support
AWS infrastructure support
AWS Support is a one-on-one, fast-response support channel that is staffed 24x7x365 with experienced and technical support engineers. The service helps customers of all sizes and technical abilities to successfully utilize the products and features provided by Amazon Web Services.
Similar products
Liquid LFM 7B is designed to handle complex tasks, offering an optimal balance between size and output quality.
Liquid LFM 40B is designed to handle complex tasks, offering an optimal balance between size and output quality.
Liquid LFM 40B is designed to handle complex tasks, offering an optimal balance between size and output quality.
A specific implementation of Liquid Sales for field sales teams to manage accounts, contacts, deals, and complex pricing, with rapid ordering powered by real-time pricing and inventory.
Customer reviews
No customer reviews yet
Be the first to review this product . We've partnered with PeerSpot to gather customer feedback. You can share your experience by writing or recording a review, or scheduling a call with a PeerSpot analyst.