Recently, OpenAI announced the open-sourcing of the GPT-OSS-20B and GPT-OSS-120B models—large language models specifically trained for open-source purposes. In terms of capability, these models are said to rival others in the o3/o4-mini family.
Previously, Elon Musk had pledged to open source the xAI Grok 2 model. However, after the release of Grok 3, Musk failed to fulfill that promise, drawing criticism from the community. Even with the subsequent launch of Grok 4, Grok 2 remains unreleased.
According to Musk’s latest post on X, xAI now plans to open source the Grok 2 model next week. Originally released in August 2024, the model is said to offer performance comparable to Anthropic’s Claude 3.5 Sonnet and OpenAI’s GPT-4-Turbo.
The model reportedly excels in reasoning and multimodal capabilities—particularly in handling both text and images. It leverages a Mixture of Experts (MoE) architecture to enhance generation speed and computational efficiency. Once open sourced, developers and organizations will be able to fine-tune and adapt the model for specialized applications.
xAI has previously released Grok 1 in March 2024 under the permissive Apache 2.0 license, allowing free commercial use. However, despite multiple statements from Musk about continuing the open-source initiative, neither Grok 1.5 nor Grok 2 were made available—until now.
In recent months, Musk has faced growing backlash over this inconsistency. Critics have accused him of denouncing OpenAI’s open-source efforts while failing to uphold his own promises. It appears the release of OpenAI’s GPT-OSS models has reignited competitive pressure, prompting Musk to finally move forward with open-sourcing Grok 2.
Related Posts:
- Critical Sandbox Escape Vulnerabilities in VM2 Library
- xAI’s Grok 3 Debut: Musk Challenges OpenAI
- X Blocks AI Training: Musk’s New API Rules & Grok’s Edge
- Grok 4 Unleashed: xAI’s Next-Gen AI Model Takes on OpenAI & Google
Support Our Threat Intelligence
If you find our CVE report and cybersecurity news helpful, consider supporting our work.