Post

Turns out we weren't done for major LLM releases in 2024 after all... Alibaba's Qwen just released QvQ, a "visual reasoning model" - the same chain-of-thought trick as OpenAI's o1 applied to running a prompt against an image Trying it out is a lot of fun: simonwillison.net/2024/Dec/24/...

Here's what it said when I asked it to count those pelicans

Here are all of experiments with full transcripts gist.github.com/simonw/6c296...

... and since the year still isn't over, DeepSeek published a new 685B 700+GB model on Hugging Face, no documentation yet, official announcement looks like it will come tomorrow huggingface.co/deepseek-ai/...

Here are my collected notes on DeepSeek v3 so far: simonwillison.net/2024/Dec/25/...

definitely a chinese model

On a number of predictably "sensitive" topics, the model first generates and then abruptly deletes already generated text to say "beyond scope" or something.

Wonder if that’s just their safety settings or something? I read some people ran 2.5 locally and it wasn’t as censored as the Deepseek site.

From discussion on the OpenRouter Discord it looks like this model is already available via DeepSeek API?

Yeah, Paul Gauthier used that to update his Aider polyglot benchmark already, the new DeepSeek beats Claude 3.5 Sonnet! aider.chat/docs/leaderb...