Top stories today:

1. Google rolls out Gemini 2.5 Deep Think, its top reasoning model
2. OpenAI raises $8.3B at $300B valuation as part of $40B funding plan
3. Manus unveils “Wide Research” to tackle complex tasks
4. Mistral said to seek $1B raise at $10B valuation
5. Tesla found partly liable in 2019 fatal crash, ordered to pay $243M

0. Data and calendar

Listen to our AI-generated podcast summarizing today’s newsletter (beware of hallucinations):

1. Google rolls out Gemini 2.5 Deep Think, its top reasoning model

Gemini 2.5 Deep Think outperformed Gemini 2.5 Pro, OpenAI’s o3, and xAI’s Grok 4, and its variant also earned a gold medal at IMO 2025. However, o3-Pro and Grok 4 Heavy are still the top AI models across key benchmarks:

• Gemini Ultra subscribers can use Gemini 2.5 Deep Think.

• May: Google unveiled the new model at its I/O 2025.

• Can consider multiple ideas simultaneously using a multi-agent architecture.

2. Update: OpenAI raises $8.3B at $300B valuation as part of $40B funding plan

• Dragoneer Investment has emerged as the biggest investor in this funding, with a $2.8B commitment.

• Blackstone, TPG, T. Rowe Price, and Fidelity Management have also backed the new round.

• Founders Fund, Sequoia Capital, a16z, Coatue, Altimeter Capital, D1 Capital Partners, Tiger Global, and Thrive Capital have also participated.

• 5x oversubscribed, the round left some existing investors frustrated by limited allocations.

• Mar.: The $40B funding plan was announced, with SoftBank as the lead investor.

4. Manus unveils “Wide Research” to tackle complex tasks with parallel AI agents