Grok-2

xAI's Grok-2 represents a significant advancement in language model technology, offering state-of-the-art reasoning capabilities. As a successor to Grok-1.5, Grok-2 demonstrates substantial improvements in chat, coding, and reasoning. Alongside Grok-2, xAI is introducing Grok-2 mini, a smaller model that balances speed and quality.

Key Features of Grok-2

• Superior Performance: Grok-2 outperforms both Claude 3.5 Sonnet and GPT-4-Turbo on the LMSYS leaderboard. An early version of Grok-2 was tested under the name "sus-column-r".

• Enhanced Reasoning: Grok-2 showcases improvements in reasoning with retrieved content and tool use. It can accurately identify missing information, reason through sequences of events, and disregard irrelevant posts.

• Benchmark Results: Grok-2 demonstrates competitive performance in graduate-level science knowledge (GPQA), general knowledge (MMLU, MMLU-Pro), and math competition problems (MATH). It also excels in vision-based tasks, achieving state-of-the-art results in visual math reasoning (MathVista) and document-based question answering (DocVQA)

Additional Information

• Multimodal Understanding: xAI plans to release a preview of multimodal understanding as a core feature of Grok on 𝕏 and via the API.

• Continued Innovation: xAI is focused on advancing core reasoning capabilities and has plans for further developments.