RESEARCH·Nov 11, 2026, 9:05 PM ET·Emerging signal

Salesforce uses SFR-AutoR&D to improve embedding quality and speed

TL;DR:

  • Salesforce AI Research used SFR-AutoR&D to develop a model from EmbeddingGemma 2.
  • Salesforce AI Research measured better quality and lower latency in the resulting SFR-Embedding v4.

Summary

Salesforce AI Research used SFR-AutoR&D to improve the open-source EmbeddingGemma 2 model. The resulting model, SFR-Embedding v4, showed gains in quality and latency. The supplied details don't quantify those improvements or establish how much research effort the agents saved.

Open salesforce.com in a new tab ↗
Powered by Moatly Lens

The Bigger Picture

The headline tells you what happened. Discover the facts behind it, the forces at work, the connections that matter, and what comes next.

Why This Matters

Salesforce is using its own AI agents to build better AI models, resulting in a new version of its embedding technology.

What We Actually Know

  • Reported, not yet confirmedThe development process utilized a system called SFR-AutoR&D to refine an existing open-source model.

How the Pieces Connect

  • InferenceReported, not yet confirmedThis project demonstrates the practical use of 'AI for AI' where autonomous agents handle the iterative research tasks usually done by humans.

What This Could Change

  • ScenarioReported, not yet confirmedIf these agents can consistently improve open-source models, the speed of AI model iteration could increase significantly across the industry.

Where the Story Goes Next

  • How much human intervention was actually required during the SFR-AutoR&D process?