Thursday 9 July 2026
Top stories today:
Grok 4.5 launches as SpaceXAI’s first AI model built with Cursor
GPT-Live launches as OpenAI’s new full-duplex voice models
OpenAI says ~30% of SWE-Bench Pro tasks are broken
Blue Origin to raise $10B at $130B pre-money valuation
Wonder, GrubHub parent, to raise millions at $9B valuation
0. Data & Calendar
All values as of 6 AM ET / 3 AM PT, other than S&P500 and NASDAQ close (4 PM ET / 1 PM PT).
All times are ET.
Listen to our AI-generated podcast summarizing today’s newsletter (beware of hallucinations):
1. Grok 4.5 launches as SpaceXAI’s first AI model built with Cursor
Grok 4.5 has surpassed Claude Sonnet 5 to rank 4th on the Artificial Analysis Intelligence Index:
It costs $2/$6 per 1M input/output tokens, vs. GPT-5.5’s $5/$30, Opus 4.8 and Sonnet 5’s $3/$15, and GLM-5.2’s $1.4/$4.4.
With 1.5T parameters, it’s 3x larger than its predecessor, Elon Musk said.
Designed to “handle difficult, long-running” legal, finance, and coding tasks.
Grok 4.5 also scored on par with GPT-5.5 in Codex on the Artificial Analysis Coding Agent Index
Available in Grok Build, Cursor, and the SpaceXAI Console, but not in the EU.
Coresignal turns 10 this year, and you get the gift.
For a decade, we've been providing fresh company, employee, and jobs data for 1000+ organizations. Now's the best time for you to put this data to work and build your AI search, sales tech, or HR tech products.
Use code BDAY50 for 50% off your first month's subscription at checkout on our self-service platform.
Limited-time offer
Only valid for monthly subscriptions
New customers only
*Sponsored.
3. GPT-Live launches as OpenAI’s new full-duplex voice models
GPT-Live-1 has outperformed the existing Advanced Voice Mode on early benchmarks run by OpenAI:
GPT-Live-1 and GPT-Live-1 mini can both listen and speak simultaneously.
GPT-Live-1 is rolling out to ChatGPT Voice for Go, Plus, and Pro users, while GPT-Live-1 mini becomes the default for free users.
HR and IT need to work as one. Here's how
HR and IT misalignment costs you more than time. Get the guide that closes the gaps across the full employee lifecycle.
5. OpenAI says ~30% of SWE-Bench Pro tasks are broken, retracts recommendation
OpenAI said that it found 27%–34% of SWE-Bench Pro tasks to be broken in its analysis:
An automated filter flagged 286 potentially broken tasks for further human review.
OpenAI created a quality assurance pipeline to assess benchmark reliability
Fill key roles faster, without waiting weeks or months for the pipeline to behave.
AI-assisted matching and structured vetting help you find people who fit the role and the team.
Cut time and cost with a process built for scale, not endless back-and-forth.
*Sponsored.
