Google Unveils Data Science Agent & DiffRhythm Generates Songs in Minutes
AI Model Releases
Google Launches Data Science Agent: Google has introduced a new Data Science Agent integrated with Colab and Gemini to boost data science capabilities. This tool is predicted to revolutionize the data science field, possibly making certain jobs redundant while enhancing economic productivity.
Grok Preview Outperforms GPT-4.5: The Grok preview model has surpassed GPT-4.5 by one point on the lmarena benchmark, particularly excelling in the 'Hard Prompts w/ Style Control' metric.
DiffRhythm Released by ASLP-lab: ASLP-lab has unveiled DiffRhythm, an AI model that can generate full-length songs with vocals in just 4 minutes using latent diffusion.
CohereForAI Introduces C4AI Aya Vision: CohereForAI has released C4AI Aya Vision, a 32B parameter model being compared to Qwen2.5 72B.
AI Benchmarking
ByteDance Releases SuperGPQA Benchmark: ByteDance, along with the M-A-P open-source community, has introduced SuperGPQA, a new benchmark for evaluating large language models across 285 graduate-level disciplines.
GPT 4.5 Dominates lmarena Benchmark: GPT 4.5 has achieved the highest scores in all categories of the lmarena benchmark, outperforming models like Gemini and Sonnet.
AI Hardware and Products
Perplexity Developing AI Phone: Perplexity is collaborating with Deutsche Telekom to create a new AI Phone featuring an app-less design and voice command functionality, set to launch in late 2025.