Build Your Own LLMs Personalized. Private. Affordable.
The ThirdAI engine makes it easy to build and deploy billion parameter models on just CPUs.
No configs. No GPUs. No latency.
What can you do with Large Language Models?
import thirdai
load_data(“wikipedia.json”)
train_model(“search_wiki”)
print_answer(“is everest in india”)
The Third AI Difference: Pre-train on your data instead of just using public models
| Sentiment Analysis |
|---|
| Method | RoBERTa (fine tuned for sentiment) | ThirdAI Bolt |
|---|---|---|
|
Accuracy |
83.02% |
93% |
|
Training Time |
40 hours on GPU |
20 min on laptop CPU |
|
Inference latency (ms) |
46 |
1 |
| SciFact Benchmark |
|---|
| Method | T5-Large | ThirdAI UDT |
|---|---|---|
|
Precision@1 |
39 |
58 |
|
Recall@100 |
82 |
90 |
|
|
|
|
|
|
|
|
|
|
|
|
| Information Retrieval on MSMarco |
|---|
| Method | ColBERT V2 | ThirdAI BOLT |
|---|---|---|
|
Latency (ms) |
721 |
100 |
|
Recall |
.965 |
.962 |
|
|
|
|
Unified Interference
Our Universal Deep Transformers (UDT) library for AutoML can tackle a broad range of machine learning tasks and data modalities all through the same unified API.
Dramatically Lower Cost of Model Training and Refreshing
Improve Inference by 100X or More
Our proprietary sparse inference algorithms and Universal Deep Transformer library can achieve inference latencies under 1 millisecond on standard CPUs.
Immediately Production-Ready
Explainable/Trustable AI
ThirdAI’s production-ready UDT software is inherently designed to provide attribution and explanation for any AI decisions made.
Sustainable/Green AI
Our UDT can help achieve significantly lower carbon emissions for some of the most complex AI workloads without losing on the accuracy.
IN THE NEWS
How Wayfair’s scientists collaborated with innovative startup ThirdAI to serve hyper-relevant search results to customers
ThirdAI founder works to make artificial intelligence more efficient
Got CPU? Unlocking AI Speed and Flexibility
The current AI industry assumes that CPUs are inferior to GPUs and other specialized processors (such as TPUs) for heavyweight AI computations.
