World's most calming AI chatbot (imo), Pi by Inflection, gets powerful IQ to already exceptional EQ
2 min. read
Published on
Share this article
Improve this guide
Read our disclosure page to find out how can you help MSPoweruser sustain the editorial team Read more
Key notes
- Inflection-2.5 is a powerful upgrade to the personal AI assistant Pi, achieving high performance with less computing power.
- Pi users benefit from improved performance on tasks like coding, math, and web searches.
- Inflection-2.5 demonstrates strong results in STEM areas, scoring high on benchmarks.
- Inflection-2.5 maintains Pi’s focus on empathy and safety while offering increased capabilities.
Inflection-2.5, a new upgrade to the personal AI assistant and my personal favorite, Pi, gets new improvements in performance while requiring less computing power. The last released Inflection-2, which then was claimed to be the best LLM back then.
Previously known for its focus on empathy and safety, Pi now incorporates powerful IQ capabilities on par with LLMs like GPT-4. Despite using only 40% of the training resources required by GPT-4, Inflection-2.5 achieves over 94% of its performance on various benchmarks.
Benchmarks show greater gains in STEM (Science, Technology, Engineering, and Mathematics) areas, with Inflection-2.5 scoring impressively on the Hungarian Math exam and the Physics GRE.
The company also acknowledges and addresses flaws identified in a widely used benchmark dataset.
“After evaluating MT-Bench, we realized that a large fraction—nearly 25%—of examples in the reasoning, math, and coding categories had incorrect reference solutions or questions with flawed premises. “
Overall, Inflection-2.5 marks a significant step forward for Pi. By combining high IQ with its existing strengths in empathy and safety, Pi is poised to become an even more helpful companion for millions of users worldwide.
“All evaluations above are done with the model that is now powering Pi, however we note that the user experience may be slightly different due to the impact of web retrieval”
More here.
User forum
0 messages