GPT-5.4 nano is also available starting today in the API. [Translated from EN to English]
LLMS
-

OpenAI Launches GPT-5.4 mini, Twice as Fast
By
–
GPT-5.4 mini is available today in ChatGPT, Codex, and the API. Optimized for coding, computer use, multimodal understanding, and subagents. And it's 2x faster than GPT-5 mini. https://openai.com/index/introducing-gpt-5-4-mini-and-nano/ [Translated from EN to English]
-

Princeton Review Highlights Persistent LLM Reliability Issues
By
–
BREAKING: Reliability, which I have been harping on here since 2019, continues to be deep problem, even with the latest models. A new @Princeton review below offers a taxonomy of some of the many ways in which reliability continues to haunt LLMs seven years and a trillion
-

Open Models Panel at GTC: Industry Leaders Discuss Future
By
–
Open Models Panel at GTC with Harrison & Jensen: Join us tomorrow, Wednesday March 18th at 12:30pm at GTC for “Open Models: Where We Are and Where We’re Headed”, a panel featuring Harrison, Jensen, and the CEOs of Cursor, Thinking Machines Lab, Perplexity, and more. Add it to
-
Framework-specific challenges with VRAM, disk, and quantization options
By
–
Thanks for the feedback. The only problem with VRAM, disk, and quant options is that they are framework specific. Disk is maybe the most obvious or feasible one for a given precision (like bf16)
-
LLM Slop: Repetitive Words Across Different Contexts
By
–
one thing people miss about LLM slop is the slop is NOT in the actual words. it’s that the same words get used over and over in very different contexts. anything can be slop
-
Inference VRAM Complexity Depends on Framework Implementation
By
–
I was thinking about inference VRAM as well, but this one is so tricky because it depends on the implementation/framework one is using
-
Elon Musk asks Grok to vulgarly roast other AIs
By
–
@grok please do a vulgar roast of the other AIs in the voice of those AIs
-

Grok 4.2 Multi-Agent Token Efficiency Benchmark Update
By
–
That reasoning basically doesn't help at all, the only benchmark I can think of that shows this. Maybe I need to update this chart as Grok 4.2 multi agent ate up way too many tokens
-
The Sycophancy Problem in AI Models: Why Flattery Signals Useless Answers
By
–
The sycophancy problem is so real haha. Every time a model starts with "What a great question!" I already know the answer is going to be useless.