Sold by: Trillion Labs
Deployed on AWS
A 16B Mixture-of-Experts reasoning LLM (about 3B active) with step-by-step reasoning output, 32K context, and an OpenAI-compatible chat API.
Overview
Gravity-16B-A3B-Preview is a Mixture-of-Experts (MoE) large language model from Trillion Labs, built to deliver strong reasoning and generation quality at a fraction of the inference cost of comparable dense models. It has 16 billion total parameters but activates only about 3 billion per token, routing each token through a small subset of expert networks. This keeps latency and GPU cost low while preserving the breadth of a much larger model.
The model generates explicit step-by-step reasoning before its final answer, which improves reliability on analytical and multi-step tasks, and it supports a 32,768-token context window for long documents and conversations. It performs strongly in both English and Korean, making it well suited to multilingual and Korea-focused applications.
Gravity exposes a standard OpenAI-compatible chat-completions API, so existing OpenAI client libraries and tooling work without code changes. Deploy it as a real-time Amazon SageMaker endpoint for interactive applications, or run batch transform jobs to process large datasets offline. Target users: developers and enterprises building reasoning, chat, agentic, or content-generation applications, including bilingual English and Korean use cases.
Highlights
- Cost-efficient MoE: only about 3B of 16B parameters activate per token, delivering high-quality output at lower GPU cost and latency than dense models of comparable capability.
- Built-in reasoning: produces explicit step-by-step reasoning before its final answer, improving reliability on multi-step and analytical tasks.
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
This product is available free of charge. Free 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.
Dimensions summary
This listing carries no software charges. You pay only for the AWS compute instance you run it on. Pricing is organized by instance type and inference mode, billed per host hour. One option uses the ml.g5.12xlarge in batch mode, which processes grouped requests. The rest run in real-time mode, which responds to requests as they arrive. Real-time choices span several instance families and sizes, from the ml.g6e.xlarge up to the ml.g5.48xlarge. Larger instances provide more compute per hour. Choose the instance that matches your workload size and response needs.
Top-of-mind questions for buyers
One host hour is one hour that a single AWS instance runs this model. You are charged per running instance, per hour. If you launch two instances, both accrue host hours at the same time. Billing tracks running time, not the number of requests processed.
The model itself carries no software charge, so no software fee accrues. Host hours meter running time only, so a fully stopped instance stops accruing them. Stopped instances may still incur underlying AWS charges, such as storage, billed separately by AWS.
Batch mode is offered on the ml.g5.12xlarge and processes grouped requests together. Real-time mode responds to requests as they arrive and is available across the other instance types. Both bill per host hour, so cost tracks how long each instance runs, not the mode you pick.
trillionlabs.co
Helpful?
Vendor refund policy
Software charges for Gravity-16B-A3B-Preview are metered by AWS Marketplace based on your usage (per instance-hour, or per the terms of your offer) and are generally non-refundable once incurred.
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
Tool calling enabled
Additional details
Inputs
- Summary
POST to /invocations with Content-Type application/json. Body is an OpenAI chat-completions object: model (must be Gravity-16B-A3B-Preview), messages array (roles system, user, assistant), optional max_tokens, temperature, top_p, stop.
- Input MIME type
- application/json
{
"model": "Gravity-16B-A3B-Preview",
"messages": [
{"role": "system", "content": "You are a helpful assistant."},
{"role": "user", "content": "What is the capital of South Korea? Answer in one sentence."}
],
"max_tokens": 512,
"temperature": 0.7,
"top_p": 0.9
}
{"model":"Gravity-16B-A3B-Preview","messages":[{"role":"user","content":"Summarize: ..."}],"max_tokens":256}
{"model":"Gravity-16B-A3B-Preview","messages":[{"role":"user","content":"Translate to Korean: Hello"}],"max_tokens":128}
Support
Vendor 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.
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.