Many users criticized Gemini 3.6 Flash as uncompetitive and underwhelming compared to GPT-5.6 Sol because it costs more despite inferior speed and intelligence.
Based on 15 visible X reactions from 28 accounts; directional sample.
Ask a question below.
Published answers will appear here.
The OpenAI model costs $0.31 per benchmark task.
@theo its even worse for 3.5-flash-lite the model thats supposed to be dirt cheap compared with gpt-5.6-luna-medium it: - scores worse - uses more tokens - costs more - is 2x slower to respond there is genuinely no metric they dont lose on thats a quadruple whammy
Google's in a bind. Talent exodus. Leading model is underwhelming and can't be released. Success in the AI arena means more disruption to their ad-fueled revenue model. They have a fair bit of challenges ahead of them. Maybe Kimi3 could help them think through their options...
@theo Yeah, the new Flash isn't good. I put it down after 30 min and it putting what I asked it to work on in a Git-ignore folder that I only use for configs.
@theo I鈥檇 really like to know what Google鈥檚 product strategy here is for releasing a model that鈥檚 so uncompetitive that even the open source ones outperform it
Many users criticized Gemini 3.6 Flash as uncompetitive and underwhelming compared to GPT-5.6 Sol because it costs more despite inferior speed and intelligence.
Based on 15 visible X reactions from 28 accounts; directional sample.
Ask a question below.
Published answers will appear here.
Measuring a model's speed by TPS is like measuring a bike's speed by RPM. Sounds smart until someone shows up with a tricycle
Honestly 馃槶