Introduction to AI model tests
I tested two new AI models for a whole day. GPT 5.6 Sol from OpenAI and Fable 5 from Anthropic. I wanted to see which one works better in daily work.
Many people only look at test results. But I preferred to test both models myself. I did several different tasks and compared the results.
At AI w Biznesie we know that choosing a model matters a lot. Each task needs different tools. That is why real-world tests give the best picture.
First impressions and benchmark tests
On paper both models look very good. GPT 5.6 Sol has high scores in coding tests. Fable 5 is better at more complex programming tasks.
In the first test I asked both models to make a browser game. It had to be an open world with bicycles. The game had to work well and be fun.
Fable 5 made a 3D game from a first-person view. The map was big and the bike physics worked great. The game looked like a small GTA.
GPT 5.6 Sol made a game from a top-down view. It was functional but harder to control. It looked less interesting than the game from Fable.
The time difference was small. Fable made the game in 21 minutes and Sol in 23 minutes. But the cost was very different. Fable cost 14 dollars and Sol only 4.50 dollars.
At AI w Biznesie we often see such situations. A cheaper model can be good but does not always give better quality. In this test Fable won on creativity.
Interactive website
The second test was to make an impressive website. The site had to stop the user and show an interesting story. Both models had full freedom.
Fable 5 created a site about 10 billion years of history. It was full of 3D effects and sounds. The site reacted to mouse movements and looked amazing.
GPT 5.6 Sol made a similar site but less polished. It had nice elements but lacked the „wow effect”. Fable again was better visually.
The cost difference was huge. Fable cost 19 dollars and worked for 23 minutes. Sol cost only 1 dollar and worked for 7 minutes. Sol was 20 times cheaper.
Is it worth paying 20 times more for better quality? It depends on the task and budget. At AI w Biznesie we always advise matching the tool to the need.
More complex tasks and API tests
The third test was more ambitious. I asked the models to make five different visual elements. They could be games, sites, simulations or something else.
Fable 5 made a gallery of five worlds. There were flight simulations, arcade games and interactive stories. Unfortunately some elements had bugs.
GPT 5.6 Sol made a gallery called „Impossible Objects”. It showed more variety of ideas. There were simulations, word games and presentations. Sol won this test.
Sol was also faster and cheaper. It worked for 7 minutes and cost 1.22 dollars. Fable needed 15 minutes and cost 15 dollars. Sol was 12 times cheaper.
At AI w Biznesie we teach clients that different tasks need different models. For quick idea generation Sol is great. For creative quality Fable is better.
Quick API tasks
I also did a series of quick API queries. Each question was simple and needed one answer. I checked speed and cost of both models.
The results were very interesting. Sol won 24 times and Fable only 3 times. Fable often refused to answer because of strict safety rules. This lowered its effectiveness.
The cost was also different. Sol cost 16 dollars and Fable 63 dollars. When Fable answered, the quality was similar to Sol. The problem was reliability.
At AI w Biznesie we know that reliability is very important. A model that often refuses loses value. Sol proved more practical for simple tasks.
Key differences between the models
After all tests I can clearly say what makes these models different. Fable 5 is like a manager or co-founder of a company. GPT 5.6 Sol is like a good worker.
Fable is great at strategic thinking. It is creative and gives good advice. It writes better and creates videos better. But it is more expensive and slower.
Sol does tasks quickly. It is cheap and efficient. It rarely refuses to answer. It works well in tests and coding. But it is less creative.
At AI w Biznesie we often talk about matching tools. If you need strategy, choose Fable. If you need fast execution, choose Sol.
The cost difference is big. Fable costs 10 dollars per million input tokens. Sol costs 5 dollars. For output Fable costs 50 dollars and Sol costs 30 dollars.
Sol is also more efficient with tokens. It often uses half the tokens that Fable uses. This means even bigger savings.
Verdict and recommendations
After a full day of testing I have a clear opinion. Fable 5 is a better model in terms of ability. It is more creative and strategic. It is for demanding tasks.
GPT 5.6 Sol is a better model in terms of economy. It is cheaper, faster and more efficient. It is for daily work and quick tasks.
The ideal strategy is to use Fable as a manager. It can control a group of Sol agents. Fable makes decisions and Sol executes them quickly.
At AI w Biznesie we recommend this approach. Combining the strengths of different models gives the best results. Clients save time and money.
When to choose Fable 5
Choose Fable 5 if you need a generally available model. Sol is in a limited preview version. Fable works right away all over the world.
Choose Fable if creativity and strategy are your priority. It is better at writing, advising and creating videos. It works well in complex projects.
Choose Fable if you can accept a higher cost. The manager quality is worth its price. At AI w Biznesie we use it for important decisions.
When to choose GPT 5.6 Sol
Choose Sol if you need the best cost performance. It is cheaper and more efficient with tokens. It works well in budget projects.
Choose Sol if fast execution is your priority. It is faster in coding and tests. It refuses to answer less often, which increases reliability.
Choose Sol if you have access to the preview version. At AI w Biznesie we use it for daily tasks. It gives great results at a low price.
To sum up, both models are very good. Fable 5 is in a different class for creativity. Sol 5.6 is more comparable to the older Opus 4.8.
If OpenAI wants to compete with Fable, it must release a GPT 6 model. For now Sol is a great but cheaper choice. At AI w Biznesie we follow these changes and advise clients on the best solutions.
No responses yet