Results

The Center for AI Standards and Innovation (CAISI) evaluated the open-weight AI model DeepSeek V4 Pro and found it lags frontier models by about 8 months.

CAISI, a unit of NIST, has a full report, but the most notable comparison is cost. After all, a model that lags frontier models by 8 months may be just swell for most use cases.

"CAISI selected a U.S. reference model by filtering out U.S. models that performed significantly worse on public benchmarks or that cost significantly more per token than DeepSeek V4 Pro. The only model meeting these criteria was GPT-5.4 mini, which was selected as a point of reference. In CAISI’s aggregated capability analysis, GPT-5.4 mini receives an Elo score of 749, which is similar to DeepSeek V4 Pro’s score of 800.

DeepSeek V4 costs less than GPT-5.4 mini on 5 out of 7 CAISI benchmarks."

DeepSeek V4 Pro

Ever since Microsoft reported its fiscal third quarter earnings I've been stuck on one word: Eval-max. Eval-max? What?!?

Microsoft CEO Satya Nadella said the following (emphasis mine):

"It was a record third quarter powered by the continued strength of Microsoft Cloud, which exceeded $54 billion in revenue, up 29% year-over-year. Our AI business surpassed $37 billion ARR, up 123%. We are at the beginning of one of the most consequential platform shifts that will change the entire tech stack as agents proliferate and become the dominant workload. This will drive TAM expansion and change the value creation equation across the entire economy. To capture this opportunity, we are executing against 2 priorities. First, we are building the world's leading cloud and AI infrastructure for agentic computing era. Second, we are building high-value agentic systems across core domains such as productivity, coding and security. These 2 layers reinforce each other, and we are focused on driving competitive value and differentiation for customers across each so that they can eval-max their outcomes."

Max could become the new "super." All you have to do is attach it to every word and you'll sound ___. I'll leave the blank to you.

Earnings to know after the bell.

  • Storage remains hot as SanDisk and Western Digital topped estimates. SanDisk's fiscal third quarter earnings of $23.03 a share crushed estimates on revenue of $5.95 billion, up 97% from a year ago. Data center revenue was up 233% from a year ago. Fourth quarter revenue will be between $7.75 billion to $8.25 billion.
  • Western Digital's third quarter also shined with earnings of $8.20 a share on revenue of $3.34 billion, up 45% from a year ago. Fourth quarter revenue will be about $3.65 billion.
  • Atlassian eased SaaS fears with better than expected third quarter results. The company reported a third quarter net loss of $98.4 million, or 38 cents a share, on revenue of $1.79 billion, up 32% from a year ago. Non-GAAP earnings in the third quarter were $1.75 a share. Atlassian projected fourth quarter revenue of $1.653 million to $1.661 million. Shares surged in afterhours trading.
  • Rivian's first quarter was notable not because of its vehicle production (10,236), but because its software and services revenue was $473 million out of a total of $1.38 billion. Software and services revenue was up 49% from a year ago courtesy of vehicle electrical architecture and software development services as well as vehicle repair and maintenance. Rivian counts VW as a big investor.

Meta released its first quarter results and overall they were strong with revenue up 33% from a year ago. As usual, Meta said it will spend more money on AI infrastructure. The company said 2026 capital expenditures will be between $125 billion to $145 billion, up from the $115 billion to $135 billion forecast last quarter.

Here's the problem. Unlike Google, Amazon and Microsoft, Meta has to monetize AI through advertising. It doesn't have a cloud business where it can sell capacity to enterprises.

Meta Q1 2026 Capex