OPENAI NO YOU CAN'T LET GOOGLE WIN LIKE THIS

I will miss the funny model

"spark can you summarize this article for me" :

Luna Spark announcement imminent, I think

You’re too naive Ben, this needs special sauce to pull off. You can’t just be ultrafast You need to be ultrastupid too

ahahahahah

Google: 3.8 - OpenAI: 0

Codex-spark is the dumbest model ever released by openai

The demand for these new models is actually causing some massive growing pains for their infrastructure.

i cant tell if i like or dislike these bouncy charts

I think it'll be replaced with 5.6 sol or possibly 6 sol

What is the actual measurement here? As reading a file is going to happen client side so it cannot be measuring what it says it is

you almost spit my coffee out on my keyboard.

This was a great idea btw

It’s not winning if the speed results in poor quality

im less concerned with speed than with the accuracy and precision I receive from GPT or Codex. GPT Spark proved costly and careless because it had to operate at those speeds - the trade-offs it required and the context shifts needed to sustain that on anything more complex than processing an Excel file were exhausting - a sunset on the dawn of astra is a good retirement plan

Does that mean Spark is now your favorite model?

Try Qwen on cerebras at 1500 tps

the balls benchmark

fr they literally said they have better models now. this is not the L you think it is

pointless model atp