Inference performance on Huawei 910C achieves 60% of the H100's performance (?)
O3 mini new king of Coding.
Oh my god
Why people are really underestimating Google
o3 vs R1 on benchmarks
I don't need another flash model
I feel like their release schedule disturbed
Anthropic CEO says blocking AI chips to China is of existential importance after DeepSeeks release in new blog post.
C'mon Claude, do something
How many people here think that Google will overtake ChatGPT?
NVDA is doomed if deepseek is legit
DeepSeek officially tops the AppStore
Can we give Google some credit for currently having the only model in the top 5 without "thinking"? Their full thinking model is probably going to be #1 once its finished testing.
IMO Deepseek is the first model that gives a good challenge to Claude
The Music
Comparing AI Search Engines: SearchGPT, Perplexity, and Claude
I think it's forced. DeepSeek did its best...
Testing Claude, ChatGPT and Gemini for medical image analysis: brain anatomy
Rumors of industry panic caused by DeepSeek
Livebench results updated for gemini-2.0-flash-thinking-exp-01-21
Open-source Deepseek beat not so OpenAI in 'humanity's last exam' !
OpenAI and SoftBank will each invest $19B into Stargate
Humanity's Last Exam dataset is out!
Deepseek R1 thinks more naturally, more elaborately, and more sophisticatedly than Flash thinking. And it's so much better at code too.
Google's Gemini 2.0 Flash Thinking Exp 01-21 model now has a context window of over 1M tokens.
