Atlas

Models

← All models

Grok 2 Vision 1212

xAI · 2024-12-12

Grok 2 Vision 1212 is xAI's December 2024 multimodal snapshot of Grok 2, optimized for improved accuracy, instruction following, and multilingual support across both text and image inputs. It is a production-ready vision-language model capable of processing documents, diagrams, photographs, and other visual inputs alongside text. The 1212 date suffix indicates the December 12, 2024 checkpoint used for this stable API release.

Benchmark scores

BenchmarkScore
AI2D (OpenVLM)81.2
Capability123.1
HallusionBench (OpenVLM)51.7
MathVista (OpenVLM)66.6
MM-Vet (OpenVLM)65.2
MMMU Validation (OpenVLM)67.1
MMMU-Pro (Vals 4-Option)57.3
MMStar (OpenVLM)65.2
MortgageTax - Numerical Extraction (Vals all 1,258 documents)4.5
MortgageTax - Semantic Extraction (Vals all 1,258 documents)2.4
MortgageTax (Vals all 1,258 documents)3.5
Loading Atlas data…