AI Digest — July 9, 2026
Quick Notes
- SpaceXAI launched Grok 4.5 as its first Opus-class model focused on coding and agents, priced at $2/1M input tokens with a 500k context window, and is now available via multiple platforms including Hermes Agent @https://www.latent.space/p/ainews-spacexai-launches-grok-45
- Cursor partnered with xAI to train Grok 4.5, calling it their most powerful model yet and announcing in-product availability with double usage for the first week along with support from Hermes Agent, Grok Build/API, OpenRouter, and various subscription platforms @https://www.latent.space/p/ainews-spacexai-launches-grok-45
Structured Summaries
New Models / Research
xAI has launched Grok 4.5, positioning it as an “Opus-class” model specifically trained for coding workloads and autonomous agent workflows in partnership with Cursor. At approximately 1.5 trillion parameters (3x larger than the preceding Grok 4.3), the new model delivers frontier-level intelligence at competitive economics: $2 per million input tokens and $6 per million output tokens, with cache hits discounted by 75% to $0.5/1M tokens. Despite a modest 500k context window in its initial release (with Elon Musk indicating an upgrade back to 1M context “by next week”), Grok 4.5 performs comparably to the previous Opus benchmark leader even on SWE-Bench Pro, which some observers note is increasingly saturated. The model represents xAI’s strategic entry into the flagship coding-agent tier rather than a mere iterative refresh, targeting both Tesla and SpaceX engineers plus general developer adoption through multiple distribution channels including direct API access, Grok subscriptions, OpenRouter, Hermes Agent Portal, and various third-party providers.
Sources: https://www.latent.space/p/ainews-spacexai-launches-grok-45
🔗 View this digest on the web: https://ai-digest-b7u.pages.dev/digests/2026-07-09-morning/