Melbourne startup Maincode releases new AI model

Melbourne-based AI company Maincode has released a new artificial intelligence model designed to reduce the cost and infrastructure burden of AI-driven software development.

The model, Maincoder, is a compact code-generation system with one billion parameters that delivers performance comparable to much larger models on industry benchmarks, the company claims.

On the widely HumanEval test of functional code correctness, Maincoder achieved a 76 per cent score, "placing it among the top-performing open-source models of its size globally".

Maincode says the release highlights the strategic importance of efficient, locally deployable AI systems at a time when access to large-scale compute and foreign cloud infrastructure is becoming a policy and commercial concern. 

“Maincoder-1B shows the performance gains unlocked by disciplined, system-level innovation, in a field far from settled.” Dave Lemphers, founder and CEO of Maincode, said.

Maincoder can run on laptops and other constrained hardware, enabling organisations to deploy AI tools on-premises or at the edge.

The model is intended for applications such as interactive coding assistance, large-scale software maintenance, and automated code transformation, as well as being suitable for systems that require many rapid AI outputs, including testing and verification tools.

 Maincoder was developed in Melbourne by a team of researchers and engineers and has been released under the Apache 2.0 open-source licence.

The company cautions that the model is best suited to small, well-defined programming tasks and is not intended for safety-critical applications without human oversight.

The model is now publicly available via the Hugging Face platform.

Lemphers spoke to techpartner.news in November about why the cmpany chose to move away from the label of 'sovereign AI'.

“When you start wading into this sovereign conversation, it becomes infeasible because some people without actual technical capability say ‘Well, you should be going all the way down to building the silicon,’ and it's like, ‘sure but that's not actually where we need to go, it's not going to advance productivity in this country’,” he explained.