whatmodel.sh / CogVLM2 Llama3 Chat 19B

CogVLM2 Llama3 Chat 19B

+ contribute
Update model informationAdd a place to run itAdd a test resultPropose a benchmarkCorrect a benchmark scoreEdit data on GitHub

CogVLM2 Llama3 Chat 19B is an open-weights language model from Zhipu AI (THUDM), released May 2024. Its published size is 19.5B parameters. Licence: cogvlm2.

Catalogue identity notes

Comparison-only. Arena entry 'cogvlm2-llama3-chat-19b' (vision). LMArena's leaderboard table (arena-leaderboard space @6f2eeeb) labels 'cogvlm2-llama3-chat-19b' as 'CogVLM2-llama3-chat-19b' (Zhipu AI, CogVLM2), linking https://huggingface.co/THUDM/cogvlm2-llama3-chat-19B. Maker repo zai-org/cogvlm2-llama3-chat-19B @f592f29: licence tag other (cogvlm2). 19,503,107,328 parameters from the Hugging Face safetensors metadata (dense, so active = total). Released 2024-05: month the maker's Hugging Face repo was created (createdAt 2024-05-16).

43.1±2.8#480 of 719 LLMsfrom 5 benchmark results across 3 sources

$ how it compares

The LLMs ranked either side of CogVLM2 Llama3 Chat 19B, out of 828 with a score, then more from Zhipu AI (THUDM).

#modelscorewhereRAMreleased
567Hermes 2 Theta Llama 3 8B Nous Research43±2local~6.6GB2024-05
568Yi 34B Chat 01.AI43±1local~23GB2023-11
569CogVLM2 Llama3 Chat 19B Zhipu AI (THUDM)43±3local~43GB2024-05
570Code Llama 70B (base) Meta43±1local~28GB2024-01
571Gemma 3n E2B IT Google43±2local~4.7GB2025-06
more from Zhipu AI (THUDM)
349GLM-4 9B Chat 1M Zhipu AI (THUDM)53±5local~8.1GB2024-06
485CogVLM2 Llama3 Chinese Chat 19B Zhipu AI (THUDM)47±7comparison only2024-05
706ChatGLM3 6B 128K Zhipu AI (THUDM)34±4local~5.8GB2024-01

see all LLMs · find the ones that fit your machine

$ what it's good at

vision & documents▁▃▃▂▁▃▄▂▆▇█▄ #140 of 147 · 96% of its score · 5 tests

Beats Gemma 3 4B here, though they rank higher overall. Trails Qwen3.5 0.8B, Molmo 7B-D (0924), North Micro Vision Instruct (2.4B) here, though they rank lower overall.

testresultvs the others in its rundifficultycountswhy
Arena vision (style controlled)963.7 ✎▁▁▃▂▂▂▄▅▅▆██ beats 4/145avg 1068▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇ 49%blunt
CapArena-Auto detailed image captioning (Score_Avg, GPT-4o judge)-8.5000 ✎▃▃▃·▃█·▅▃▃▃▃ beats 6/12avg -0.671▇▇▇▇▇▇▇▇▇▇▇▇▇ 33%typical
ChartQA lab81.0% ✎▁··▁▁▁▁▁▂▅▅█ beats 15/44easy · avg 81%▇▇▇ 6.5%blunt · small run
OCRBench lab75.6% ✎▁·▃▃▆▅▂█▅█▁▃ beats 4/37easy · avg 82%▇▇ 5.2%blunt · small run
TextVQA lab84.2% ✎▁▁▁▄▅▁▃·▃▃▃█ beats 19/24easy · avg 78%▇ 3.0%blunt · small run

$ where to run it

Other ways to run it

version

  1. Run it as published · needs ~39.01 GB, NVIDIA only · from github.com/zai-org
    git clone https://github.com/zai-org/CogVLM2 && cd CogVLM2/basic_demo && pip install -r requirements.txt && CUDA_VISIBLE_DEVICES=0 python cli_demo.py

Weights and documentation: huggingface.co/zai-org ↗

$ how the score was worked out · LMArena leaderboard dataset 49%, CapArena: Benchmarking and An… 33%, Zhipu AI (THUDM) CogVLM2 mode… 15%

Each result implies a score on its own, after adjusting for how hard its test is and who ran it. The score is their weighted average, with a small pull toward the middle for thinly tested models. Each test's weight is its counts column above; here it's added up by who ran the tests. How it works

sourceresultsshare of the scorelicence
LMArena leaderboard dataset1▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇ 49%CC BY 4.0
CapArena: Benchmarking and Analyzing Detailed Image Captioning in the LLM Era (arXiv:2503.12329v1)1▇▇▇▇▇▇▇▇▇▇ 33%CC-BY-4.0
Zhipu AI (THUDM) CogVLM2 model card, Benchmark table (own models) · lab-reported3▇▇▇▇ 15%Factual vendor-reported score (not licensed content)
prior · pull toward the middle▇ 4%
weighted average= 43.1

$ improve this page

Update model informationAdd a place to run itAdd a test resultPropose a benchmarkCorrect a benchmark scoreEdit data on GitHub

Use a short GitHub form or edit the data through a pull request. Every number needs a source; here's which ones we can use.

data updated · sources and licences · json