AI
[AINews] SpaceXAI launches Grok 4.5, first Opus
SpaceXAI, in partnership with Cursor, has publicly launched Grok 4.5, a 1.5T parameter frontier model trained specifically for coding and agents.
![[AINews] SpaceXAI launches Grok 4.5, first Opus](https://cdn.ravenclip.com/channels/71/487/images/0-gevCrXQ1.jpg)
Key takeaways
- Grok 4.5 is SpaceXAI's first model trained specifically for coding and agents.
- Grok 4.5 is 3x larger than Grok 4.3 at 1.5T parameters.
- Grok 4.5 is roughly comparable to Opus 4.7, but much faster and more token-efficient.
- Grok 4.5 sits on the Pareto frontier for cost and performance.
SpaceXAI, in partnership with Cursor, has publicly launched Grok 4.5, a 1.5T parameter frontier model trained specifically for coding and agents. Positioned as an "Opus-class" model focusing on speed and cost efficiency rather than absolute benchmark supremacy, Grok 4.5 is priced at $2 per 1M input tokens and $6 per 1M output tokens. It features a 500k context window, which Elon Musk indicated will likely upgrade back to 1M soon. In evaluations by Artificial Analysis, Grok 4.5 ranked #4 on their Intelligence Index with a score of 54, demonstrating near-frontier capabilities with high token efficiency.
In their words
“We've partnered with SpaceXAI to train Grok 4.5. It’s our most powerful model yet and the first we've built for more than software engineering.”
By the numbers
- 1.5T
- parameter size of Grok 4.5
- 3x
- size increase of Grok 4.5 versus Grok 4.3
- $2
- cost per 1M input tokens for Grok 4.5
- $6
- cost per 1M output tokens for Grok 4.5
- #4
- rank on Artificial Analysis Intelligence Index
How it unfolded
- Cursor announces Grok 4.5 training partnership
- SpaceXAI publicly launches Grok 4.5
- GPT 5.6 is confirmed to launch
Turn stories like this into views
Ravenclip finds the AI news, makes the video, and posts it before attention moves on.
Common questions
- What happened with Grok 4.5?
- Grok 4.5 is SpaceXAI's first model trained specifically for coding and agents.
- Where can I read the original report?
- Read the full report at latent_space.
