Cogvlm 2 Llama 3 Chat 19B
CogVLM2 Llama 3 Chat 19B is a 19-billion parameter chat model built by Z.ai, based on Meta-Llama-3-8B-Instruct. It is genuinely best at image understanding tasks, showing significant improvements on benchmarks like TextVQA and DocVQA. A notable technical trait is its support for high-resolution image inputs up to 1344x1344 pixels.
Input
Output
Context
8K
Max Output
2K
Parameters
19.5B
Technical Specifications
Model TypeChat
Context Window8,192 tokens
Max Output Tokens2,048 tokens
Parameters19.5B
Release DateMay 16, 2024
Training CutoffNot available
Licenseother
Open SourceYes
Input Modalities
Text
Output Modalities
Text
Capabilities
Estimated GPU Requirements for Cogvlm 2 Llama 3 Chat 19B
Estimates based on INT8 quantization. Actual requirements vary by framework and configuration.
Related Tools
Data sourced from official provider APIs and documentation
Last updated: Jun 23, 2026
Start building with the right model.
Automatically route workloads to the right model for every task, every time.