---
format: "aidr-story-markdown/v1"
id: "7ca7cf203834173c374e35968400b3ef1a7e40bfa78074eaedff8e59a141fa88"
canonical_url: "https://aidr.today/7ca7cf20?lang=en"
title: "xAI Finishes Grok 4.8 2.5T Model Training This Week on New C++ Stack"
lang: "en"
requested_lang: "en"
available_langs: ["en","vi"]
translation_fallback: null
fallback_fields: []
published_at: "2026-09-14T02:47:02.000Z"
category: "Models"
topics: ["xai","grok","llm","inference","chips","infra","training","parameters"]
source_urls: ["https://huggingnews.com/ai/xai-finishes-grok-48-25t-model-training-this-week-on-new-c-stack-69b00172","https://x.com/elonmusk/status/2099308197802631191","https://marketbrief.now/ai/xai-finishes-grok-48-25t-model-training-this-week-on-new-c-stack-69b00172","https://huggingnews.com/ai/xai-finishes-25-trillion-parameter-grok-48-training-this-week-c791b86f","https://marketbrief.now/ai/xai-finishes-25-trillion-parameter-grok-48-training-this-week-c791b86f"]
summary: "The reinforcement learning phase for Grok 4.8 begins this week following the completion of its primary training, Elon Musk announced. The 2.5T parameter model was developed using a new C++ software stack to optimize the computational requirements and efficiency of the training process. The transition to RL occurs once the base model's initial training cycle is complete. This update confirms the specific parameter scale and the underlying software architecture for the latest iteration of the xAI assistant."
---

# xAI Finishes Grok 4\.8 2\.5T Model Training This Week on New C\+\+ Stack

> [Open the canonical story](<https://aidr.today/7ca7cf20?lang=en>)

**Published:** 2026-09-14T02:47:02.000Z
**Category:** Models
**Topics:** xai, grok, llm, inference, chips, infra, training, parameters

## Summary

The reinforcement learning phase for Grok 4\.8 begins this week following the completion of its primary training, Elon Musk announced\. The 2\.5T parameter model was developed using a new C\+\+ software stack to optimize the computational requirements and efficiency of the training process\. The transition to RL occurs once the base model's initial training cycle is complete\. This update confirms the specific parameter scale and the underlying software architecture for the latest iteration of the xAI assistant\.

## Sources

- [Story source](<https://huggingnews.com/ai/xai-finishes-grok-48-25t-model-training-this-week-on-new-c-stack-69b00172>)
- [Story source](<https://x.com/elonmusk/status/2099308197802631191>)
- [Story source](<https://marketbrief.now/ai/xai-finishes-grok-48-25t-model-training-this-week-on-new-c-stack-69b00172>)
- [Story source](<https://huggingnews.com/ai/xai-finishes-25-trillion-parameter-grok-48-training-this-week-c791b86f>)
- [Story source](<https://marketbrief.now/ai/xai-finishes-25-trillion-parameter-grok-48-training-this-week-c791b86f>)