whatmodel.sh / LLaVA-v1.6-Vicuna-13B

LLaVA-v1.6-Vicuna-13B

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

LLaVA-v1.6-Vicuna-13B is an open-weights language model from LLaVA, released January 2024. Its published size is 13.35B parameters. Licence: llama2.

Catalogue identity notes

Comparison-only, proposed so the LLaVA model-zoo runs of llava-v1-6-34b count. The pinned MODEL_ZOO.md (@c121f04) lists both own LLaVA-1.6 checkpoints in one table: liuhaotian/llava-v1.6-34b (catalogued) and liuhaotian/llava-v1.6-vicuna-13b. Weights: https://huggingface.co/liuhaotian/llava-v1.6-vicuna-13b (sha 22422b4c3a3ef1ba52aca074cc9021216877ce5d, created 2024-01-31, 13,350,844,416 params in safetensors metadata; Vicuna-13B base, so usage follows the Llama 2 licence per the model zoo note).

33.3±5.7#824 of 949 LLMsfrom 2 benchmark results across 1 source* thinly tested, treat the rank as provisional

$ how it compares

The LLMs ranked either side of LLaVA-v1.6-Vicuna-13B, out of 1132 with a score, then more from LLaVA.

#modelscorewhereRAMreleased
991Flan-T5 XXL Google33±1local~49GB2022-10
992StarCoder2 3B BigCode33±1local~3.4GB2024-02
993LLaVA-v1.6-Vicuna-13B LLaVA33±6local~9.9GB2024-01
994StripedHyena-Nous-7B Together AI33±2comparison only2023-12
995Granite 3B Code Base (2K) IBM33±1local~3.7GB2024-04
more from LLaVA
880LLaVA-v1.6-34B LLaVA40±2local~24GB2024-01

see all LLMs · find the ones that fit your machine

$ what it's good at

vision & documents▁▃▃▂▁▃▄▃▆▇█▄ #145 of 150 * · 86% of its score · 2 tests

Beats Gemma 3 4B, Phi-3-vision-128k-instruct, LLaVA-v1.6-34B here, though they rank higher overall.

testresultvs the others in its rundifficultycountswhy
MMMU (val) lab36.2% ✎▁····▃▁▄█▇▆▆ beats 1/52medium · avg 59%▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇ 74%small run
TextVQA lab67.1% ✎▁▁▁▄▅▁▃·▃▃▃█ beats 3/24easy · avg 77%▇▇▇▇▇ 12%blunt · at its level · small run

$ where to run it

Ollama · recommended

  1. Install Ollama

    download it (macOS / Windows) ↗

    Linux
    curl -fsSL https://ollama.com/install.sh | sh
  2. Run LLaVA-v1.6-Vicuna-13B · needs ~8 GB · from ollama.com
    chat in the terminal
    ollama run llava:13b
$ how the score was worked out · LLaVA repository model zoo (h… 86%

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
LLaVA repository model zoo (haotian-liu/LLaVA docs/MODEL_ZOO.md) · lab-reported2▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇ 86%Apache-2.0
prior · pull toward the middle▇▇▇▇ 14%
weighted average= 33.3

$ 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