Elfworks.ai – How we choose our AI models

Should I use the new version of ChatGPT? Is Grok actually better? I hear Claude is more accurate, is it safe? Should I try Deep Seek? What about these new ‘deep reasoning’ models?

With new versions of existing AI Models and new entrants constantly hitting the market, you need specialist knowledge to find the most appropriate model for your purposes and that of the accounting industry. We spend the time doing this testing so you don't have to! Elfworks selects the AI models that drive our Accounting Intelligence platform…

Elfworks selection process

We test each new AI model or new version of existing AI models within the following matrix:

Trust and privacy are non-negotiable; we only use enterprise grade AI models with terms and conditions that allow for no data storage and prohibit use of information to train the model. If the AI model does not meet these trust & privacy requirements we don’t use it, regardless of performance and cost.

After close to two years of continuous testing, Elfworks harnesses ChatGPT 4.1, Claude 4.0, Grok 3 and Gemini 2.5 and puts them to work on the platform. This is our “starting 4”, however we constantly test the incumbents and newcomers. If a newcomer or new version outperforms our starting 4, we switch it into our starting 4 and move the incumbent to the bench. It’s the fish that John West reject….

What is a ‘deep-reasoning’ model and are they better?

Deep reasoning models are typically not better; not if the question requires access to specialist knowledge.

If you gave a very smart person 10 seconds to answer an Australian tax question that requires specialist knowledge and you withheld that knowledge, you are likely to get an incorrect answer. If you gave a very smart person 30 seconds to answer an Australian tax question that requires specialist knowledge and you withheld that knowledge, you are likely to get an incorrect answer. Just a slower incorrect answer.

You can deep reason all you want, yet without access to the relevant specialist knowledge you are very unlikely to get to the right answer.

What is a ‘mini’ version of an AI model?

This is a smaller, more efficient version of the full AI model. They cost less to run however they compromise accuracy.

We don’t use these on the Elfworks platform as we opt for full versions of AI model that delivers the accuracy that we require to meet our selection criteria.

Contact us for free trial

The next step is to experience the benefits of an AI enhanced application created to help Australian accountants in public practice with research, advice drafting, client email responses and a wide array of productivity tools.  If you would like a free trial of the Elfworks platform, please contact us on info@elfworks.ai.

Please click here for Elfworks pricing information.