Space Exploration Technologies Corp: Grok 4.5: Delivering frontier level performance more efficiently

研报英文原文证据摘录

Space Exploration Technologies Corp: Grok 4.5: Delivering frontier level performance more efficiently

Global Research

8 July 2026ab

First Read

EquitiesSpace Exploration Technologies Corp

Grok 4.5: Delivering frontier level performance United States

more efficiently Communications Technology

12-month rating Buy

12m price target US$210.00

SpaceX releases first major Grok update with Cursor

SpaceX today released Grok 4.5, a 1.5T parameter model trained on tens of thousands

of GB300 GPUs. The release builds on Grok 4.3 (April 2026) and is the first trained in Price (07 Jul 2026) US$149.47

tandem with Cursor. Grok 4.5 is positioned to excel in coding, agentic workflows, RIC: SPCX.O BBG: SPCX US

engineering and knowledge work, delivering performance in close proximity to

Trading data and key metrics

Anthropic's Claude Opus 4.8 and OpenAI's GPT 5.5 while offering lower pricing and

greater token efficiency. The model will be priced at $2 per 1M input tokens and $6 per 52-wk range US$201.80-149.47

1M output tokens vs. $5/$25 for Opus 4.8, $10/$50 for Fable 5 and $5/30 for GPT 5.5. Market cap. US$1,967b

The press release also highlights token usage at ~16K per SWE-Bench Pro task vs. 67K Shares o/s 13,159m (COM)

for Opus, implying ~4.2x greater efficiency. Compared to Grok 4.3, the new model Free float 5%

shows sizeable improvements across agentic and coding benchmarks while continuing Avg. daily volume ('000) 172,217

to advance in knowledge and reasoning tasks. As highlighted in Figure 1, Grok 4.5 also Avg. daily value (m) US$29,568.8

narrows the performance gap relative to leading frontier models across a broad range of Common s/h equity (12/26E) US$135b

evaluations. P/BV (12/26E) 13.7x

Net debt to EBITDA (12/26E) NM

Enterprise AI: The largest long-term revenue opportunity