BREAKING : Mistral AI launched Mistral Medium 3 – a new class of models that balances:
– SOTA performance
– The model leads in professional use cases such as coding and multimodal understanding
– A range of enterprise capabilities
LLMS
-
Mistral AI Releases Mistral Medium 3 Model
By
–
-
PyTorch Foundation Welcomes vLLM and DeepSpeed Projects
By
–
the PyTorch Foundation is becoming an umbrella for great AI open-source projects. @vllm_project and @DeepSpeedAI are joining PyTorch as the first two projects! Super excited to be able to bring together multiple like-minded, high-quality projects into the foundation. Props to
-
Stanford CME 295 Transformers and LLMs Cheatsheets
By
–
Link to the cheatsheets: https://
github.com/afshinea/stanf
ord-cme-295-transformers-large-language-models
… If you're interested in ML, LLMs, and AI Agents and want to receive tutorials every week, subscribe to AI Engineering (for free): -
Stanford Releases Free LLM and Transformers Cheatsheets
By
–
Stanford University released the best cheatsheets you'll ever find to learn LLMs & Transformers! These cheatsheets cover: • Self-attention, Flash Attention, LoRA, SFT
• Mixture of Experts, Distillation, Quantization
• RAG, Agents, LLM-as-a-judge 100% Free and Open Source -
Mistral Medium 3: New Multimodal Model SOTA Performance Lower Cost
By
–
Introducing Mistral Medium 3: our new multimodal model offering SOTA performance at 8X lower cost. – A new class of models that balances performance, cost, and deployability.
– High performance in coding and function-calling.
– Full enterprise capabilities, including hybrid or -
Automating Research Tasks with AI Prompt Memorization Techniques
By
–
I’ve got strings tried around my finger to a) try adapting that to my common research tasks (I write the plan, you do the grunt work) and b) see if I can have it memorize that prompt and say “Kelsey-ize and execute: top 5 non-card payment methods in 20 largest economies, go.”
-
Verbose Strategy Prompting Improves Language Model Performance
By
–
Have you tried giving it a verbose strategy, or telling it to think of a verbose strategy then execute against the plan? @KelseyTuoc ‘s prompt for GeoGessr seems to observationally cause it to do very different things than a tweet-length prompt, which results in “winging it.”
-
Upcoming Multimodal Features and Capabilities for Grok AI
By
–
Looking at upcoming Grok features:
— 🚨 AI News | TestingCatalog (@testingcatalog) 7 mai 2025
– It could be used as a voice-first application
– It will get a new Spatial UI
– It will be optimised for 1 hand use
Should we expect Grok 3.5 to excel at real-world understanding, which would act as your assistant – see, hear and speak! https://t.co/B4IJ4Moxr2 pic.twitter.com/gHjogP9kVfLooking at upcoming Grok features:
– It could be used as a voice-first application
– It will get a new Spatial UI – It will be optimised for 1 hand use Should we expect Grok 3.5 to excel at real-world understanding, which would act as your assistant – see, hear and speak! -
LLMs Long Context Window Leverage Discussion
By
–
This one particularly leverages the long context window of llms!
-
Comparison of real-time AI voice generation capabilities
By
–
This is better than ElevenLabs for real-time stuff. Way more fluid and you actually get emotional nuance.