@corbtt @dvdcrbt can give concrete examples, but from my experience, they enable completely different things. RAG/search really just gives the model *info* to work with. Fine-tuning changes the *behavior* of the model.
RESEARCH
-
World Models: Tracking Genie 3 and LeCun’s Vision
By
–
technically zeroth: Genie 3
— swyx 🐣 (@swyx) 5 août 2025
last yr we spun up a #worldmodels channel on the @latentspacepod discord as this is obviously a thing worth tracking. someday @ylecun will be vindicated…https://t.co/GReXoN6kQatechnically zeroth: Genie 3 last yr we spun up a #worldmodels channel on the @latentspacepod discord as this is obviously a thing worth tracking. someday @ylecun will be vindicated…
-

OpenAI Open Weights Models Outperform Earlier Generation AI
By
–
The OpenAI open weights models are very impressive. These basically beat every model from eight months ago & the small one runs on a laptop. For example, when HLE came out in January, the top score was 3-4%. Been playing with the models and so far they feel like their scores.
-
Stealth AI Models Outperform Open Source Models on WebArena
By
–
Ha habido mucha especulación en las últimas semanas sobre si los modelos furtivos aparecidos en el WebArena serían los modelos open source pero al parecer no. Aquellos rendían mejor. https://t.co/tZ5SaLGSXy
— Carlos Santana (@DotCSV) 5 août 2025Ha habido mucha especulación en las últimas semanas sobre si los modelos furtivos aparecidos en el WebArena serían los modelos open source pero al parecer no. Aquellos rendían mejor.
-
OpenAI Open Source Models 120B and 20B MoE Tested
By
–
A quick test of the @OpenAI new open source models: 120b and 20b MoE models. With the @FeatureCrewPod planet generation prompt
— Peter Gostev (@petergostev) 5 août 2025
The comparison is to the best one I've seen – a yet unreleased OpenAI model, but then Sonnet 4, Kimi K2, Qwen 3 Coder and GLM-4.5 for comparison. The… pic.twitter.com/a4bh1WclKVA quick test of the @OpenAI new open source models: 120b and 20b MoE models. With the @FeatureCrewPod planet generation prompt The comparison is to the best one I've seen – a yet unreleased OpenAI model, but then Sonnet 4, Kimi K2, Qwen 3 Coder and GLM-4.5 for comparison. The
-

OpenAI Releases ‘Harmony’ Format for Building on Open-Weight Models
By
–

On top of new open-weight models, OpenAI open sourced “Harmony” format for developers to help building on top of GPT-OSS. “The format is designed to mimic the OpenAI Responses API, so if you have used that API before, this format should hopefully feel familiar to you.”
-

Confirmed Release Dates for GPT-5 and Teased Gemini 3.0 Flash
By
–

BREAKING : GPT-5 release date confirmed to be August 7 by flowith! Gemini 3.0 Flash has been teased as well but “not this week”! x.com/flowith/status…
-
Google DeepMind Announces Genie 3 World Model for Interactive Environments
By
–
Today we are announcing Genie 3, a general purpose world model by @GoogleDeepMind that can generate dynamic, interactive environments with a single text prompt.
— Google AI (@GoogleAI) 5 août 2025
World models are AI that understand facets of the world (like Veo's knowledge of intuitive physics or Genie's mastery… pic.twitter.com/JTqLIbDuy4Today we are announcing Genie 3, a general purpose world model by @GoogleDeepMind that can generate dynamic, interactive environments with a single text prompt. World models are AI that understand facets of the world (like Veo's knowledge of intuitive physics or Genie's mastery
-
Quantization and Prompt Templates Impact on Model Quality
By
–
It will be interesting to see if different quantizations and prompt templates result in different quality pelicans
-

OpenAI releases 20B and 120B open source models
By
–
If you're a developer, here's what you need to know about the new @OpenAI open source models: – They come in two sizes (20B + 120B)… both are impressive, but I'll be focusing mostly on the 120B for the rest of this post – Capabilities (at a glance) are essentially o3 level…