XGen Series by Salesforce
The XGen Series is Salesforce’s family of large language models (LLMs) developed to power generative AI solutions across various domains, including sales automation, multimodal reasoning, and actionable workflows. The series encompasses both proprietary models, such as xGen-Sales, tailored for enhancing sales operations within the Agentforce platform, and open-source models like XGen-7B and XGen-MM, which excel in conversational AI and visual-language understanding tasks.
Current Models in the XGen Series:
Proprietary Models:
xGen-Sales:
A proprietary model fine-tuned for industry-specific sales tasks.
Capabilities include generating customer insights, enriching contact lists, summarizing calls, tracking sales pipelines, and autonomously coaching representatives.
Integrated into the Agentforce platform to augment Einstein Sales Development Rep (SDR) Agent and Einstein Sales Coach Agent functionalities.
Open-Source Models:
XGen-7B (Base):
A general-purpose LLM with 7 billion parameters, available in two variants: 4K and 8K context windows.
Released under the Apache 2.0 license for commercial use.
Trained on multilingual datasets like RedPajama, Wikipedia, and Starcoder to support tasks such as conversational AI, summarization, long-form question answering, and multitask language understanding.
XGen-7B (Instruction-Tuned):
Fine-tuned on instructional datasets such as Dolly-15k and OASST1 to enhance instruction-following capabilities.
Available for research purposes only due to licensing restrictions.
XGen-MM:
A multimodal model family trained at scale on interleaved image-text data for visual-language tasks like image captioning and text-based visual question answering (VQA).
Includes variants such as xgen-mm-phi3-mini-base-r-v1 and xgen-mm-phi3-mini-instruct-r-v1.
Released under the Apache 2.0 license for commercial use.
Key Features Across the Series:
Actionable AI Tasks: Models like xGen-Sales automate complex workflows such as pipeline tracking, call summarization, and customer insights generation with high accuracy.
Multimodal Capabilities: XGen-MM models integrate text with visual inputs to enable advanced vision-language capabilities like image captioning and document analysis.
Extended Context Windows: XGen-7B supports up to 8K tokens for handling long-form conversations or complex datasets.
Open Source Accessibility: Models like XGen-7B and XGen-MM are freely available under the Apache 2.0 license for unrestricted commercial use.
Example Use Cases:
Automating lead enrichment workflows using xGen-Sales within Agentforce.
Deploying autonomous AI agents to summarize calls and track pipelines with high precision.
Enhancing customer relationship management through actionable insights generated by xGen-Sales.
Performing multimodal document analysis or visual question answering with XGen-MM models.