Top stories today:

1. Mistral launches ‘Large’ LLM, still worse than GPT-4
2. Google Gemini AI image tool to be relaunched in “next couple of weeks”
3. Zoom’s revenue +3% to $1.15B in Q4, with operating margin at 14.7%
4. Supreme Court appears skeptical of anti-content moderation laws

0. Data and calendar

1. Mistral launches ‘Large’ LLM, still worse than GPT-4, as company gets $16M investment from Microsoft

Mistral Large underperforms GPT-4 (and likely Gemini Ultra) on Mistral’s own benchmarks:

• 1+ years behind OpenAI, as GPT-4 was launched on March 14, 2023.

• Cost ~$22M to train, vs. “much more than” $50M-$100M for GPT-4.

• Could be signaling a plateau in LLM quality, as GPT-4 Turbo and Gemini Ultra have also failed to significantly improve upon GPT-4.

• Chat.mistral.ai launched in beta, with a waitlist required.

• Microsoft will provide Mistral models to its Azure customers.

• Investing $16M in Mistral, and the EU is already investigating it.

OpenAI continues to be seen as the leader in LLMs, according to prediction markets, followed by Google

2. Google Gemini AI image tool to be relaunched in “next couple of weeks”

Google Deepmind CEO Demis Hassabis said so at MWC Barcelona 2024:

• AI could bring in new devices like glasses to follow the smartphone in 5+ years, according to Hassabis.

Alphabet stock was -4.5% amid Gemini’s failure to accurately create historical images and heavy criticisms of bias

4. Zoom’s revenue +3% to $1.15B in Q4, with operating margin at 14.7%

A huge gap between GAAP and non-GAAP results remains as Zoom had $521M in stock-based compensation expense in Q4: