[whatmodel.sh](../../README.md) / LLaVA-v1.6-34B # LLaVA-v1.6-34B LLaVA · released 2024-01 · 34.8B params · apache-2.0 **score 41.0** ±1.5 · #548 of 732 · from 8 benchmark results across 5 sources Score within the [LMArena (text)](https://whatmodel.sh/?set=lmarena-text&score=set) comparison set only. ## What it's good at ● marks LLaVA-v1.6-34B among every model on that task or test, worst to best. 
task                vs the others         rank  counts  tests
vision & documents  ▁●▂▂▁▃▄▂▆▇█▄   #149 of 153     99%      8
vision & documents · #149 of 153 · 99% of its score Beats [Gemma 3 4B](gemma-3-4b.md), [Phi-3-vision-128k-instruct](phi-3-vision-128k-instruct.md) here, though they rank higher overall. Trails [Qwen3.5 0.8B](qwen3-5-0-8b.md) here, though they rank lower overall. 
test                                  result  vs the others in its run  difficulty        counts  why
CharXiv descriptive questions          51.0%  ●█▄▄▄·██·▄·█ beats 0/12   easy · avg 69%       33%  at its level
CharXiv descriptive questions (te…     51.5%  ●··▄··█··▄·▄ beats 0/6    easy · avg 69%       16%  at its level
Arena vision (style controlled)        964.4  ▁▁●▂▂▂▃▄▅▅█▇ beats 5/152  avg 1061             16%  blunt
CharXiv (reasoning questions)          22.5%  ●▂▂█▃▂·▂▂··▂ beats 0/13   hard · avg 34%       10%  typical
CapArena-Auto detailed image capt…  -67.5000  ●▃▃·▃█·▅▃▃▃▃ beats 0/12   avg -4.498           10%  typical
MMMU (val) (maker)                     51.1%  ▁····▁▁▃●▆▆▅ beats 10/53  medium · avg 59%    6.6%  at its level · small run
CharXiv reasoning questions (test…     20.0%  ●··▄▄▄█····▄ beats 0/6    hard · avg 30%      5.8%  typical
TextVQA (maker)                        69.5%  ▁▁▁▄●▁▃·▃▃▃█ beats 6/24   easy · avg 77%      0.9%  blunt · at its level · small run
## How the score was worked out Each result implies a score on its own, after adjusting for how hard that test is and who ran it. The model's 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 is added up by who ran the tests. [Method](../method.md) 
source                                    results  share of the score
CharXiv leaderboard                             2   43% ▇▇▇▇▇
CharXiv: Charting Gaps in Realistic Cha…        2   22% ▇▇▇
LMArena leaderboard dataset                     1   16% ▇▇
CapArena: Benchmarking and Analyzing De…        1   10% ▇
LLaVA repository model zoo (submitted /…        2    7% ▇
prior: pull toward the middle                        1% ▇
weighted average                                   = 41.0
Licences: CharXiv leaderboard ([Apache-2.0](https://www.apache.org/licenses/LICENSE-2.0)) · CharXiv: Charting Gaps in Realistic Chart Understanding in Multimodal LLMs ([CC BY-SA 4.0](https://creativecommons.org/licenses/by-sa/4.0/)) · LMArena leaderboard dataset ([CC BY 4.0](https://creativecommons.org/licenses/by/4.0/)) · CapArena: Benchmarking and Analyzing Detailed Image Captioning in the LLM Era ([CC-BY-4.0](https://creativecommons.org/licenses/by/4.0/)) · LLaVA repository model zoo ([Apache-2.0](https://github.com/haotian-liu/LLaVA/blob/c121f0432da27facab705978f83c4ada465e46fd/LICENSE)) ## How to run it ### Ollama (recommended) 1. Install Ollama. [Download it (macOS / Windows)](https://ollama.com/download) ```sh # Linux curl -fsSL https://ollama.com/install.sh | sh ``` 2. Run LLaVA-v1.6-34B (Q4_K_M, needs ~21.38 GB, from [ollama.com](https://ollama.com/library/llava/tags)) ```sh ollama run llava:34b-v1.6-q4_K_M ``` Other versions: Q8_0 (~42.26 GB) `ollama run llava:34b-v1.6-q8_0` Model repository and documentation: [huggingface.co/liuhaotian](https://huggingface.co/liuhaotian/llava-v1.6-34b) [interactive version](https://whatmodel.sh/model/llava-v1-6-34b) · [JSON](https://whatmodel.sh/api?format=json&all=1)