[whatmodel.sh](../../README.md) / Llama 3.2 90B Vision Instruct # Llama 3.2 90B Vision Instruct Meta · released 2024-09 · 88.59B params · llama3.2 **score 50.9** ±1.2 · #350 of 732 · from 13 benchmark results across 7 sources Score within the [LMArena (text)](https://whatmodel.sh/?set=lmarena-text&score=set) comparison set only. ## What it's good at ● marks Llama 3.2 90B Vision Instruct among every model on that task or test, worst to best. 
task                vs the others         rank  counts  tests
math                ▁▁▃▅●▆▄█▄▂▁▁   #292 of 431     38%      3
vision & documents  ▁▁▂●▁▃▄▂▆▇█▄   #131 of 153     32%      7
math · #292 of 431 · 38% of its score Beats [Molmo 2 8B](molmo-2-8b.md), [DeepSeek-Coder-V2 Instruct (0724)](deepseek-coder-v2-0724.md), [Llama 3.1 70B Instruct](llama-3-1-70b.md) here, though they rank higher overall. Trails [DeepSeek R1 Distill Qwen 1.5B](deepseek-r1-distill-qwen-1-5b.md), [Gemma 7B (base)](gemma-7b-base.md), [Mistral 7B v0.1 (base)](mistral-7b-v0-1.md) here, though they rank lower overall. 
test                      result  vs the others in its run   difficulty            counts  why
MATH level 5               39.4%  ▂▂▅▆●█▆▃▅▂▃▄ beats 31/92   hard · avg 28%           31%  sharp · at its level
OTIS mock AIME 2024–2025    2.6%  ▃●▄▃▃▃▇▇█▃▂▅ beats 22/197  very hard · avg 7.2%    4.1%  sharp · noisy
MathVista (testmini)       49.1%  ▂····▂···▂●▆ beats 3/9     medium · avg 51%        3.3%  at its level · noisy
vision & documents · #131 of 153 · 32% of its score Trails [Qwen3.5 2B](qwen3-5-2b.md), [Qwen3-VL 2B Thinking](qwen3-vl-2b-thinking.md), [Pixtral 12B (2409)](pixtral-12b.md) here, though they rank lower overall. 
test                                                         result  vs the others in its run   difficulty                counts  why
Arena vision (style controlled)                              1032.2  ▁▁▂▂●▂▃▄▅▅█▇ beats 19/152  avg 1061                     11%  blunt
CapArena-Auto detailed image capt…                         -25.6700  ▃▃▃·●█·▅▃▃▃▃ beats 3/12    avg -4.498                  6.6%  typical
MM-MT-Bench (Mistral, GPT-4o judg…  5.5000 MM-MT-Bench score (1-10)  ▅··▃·●·▃▃··▅ beats 5/9     avg 5.605 MM-MT-Bench s…    4.8%  sharp · noisy
MMMU (val)                                                    53.7%  ▂·▂······█●▄ beats 6/9     medium · avg 51%            3.6%  at its level
MathVista (testmini)                                          49.1%  ▂····▂···▂●▆ beats 3/9     medium · avg 51%            3.3%  at its level · noisy
VQAv2 (val)                                                   67.0%  ▃···●▃··▅▃▃▃ beats 1/9     easy · avg 73%              2.7%  blunt · at its level
ChartQA                                                       33.8%  ▄·▄●▄··▄▄▄▄█ beats 2/9     easy · avg 74%              0.3%  blunt · noisy
other tests (4) 
test                                result  vs the others in its run    difficulty           counts  why
GPQA Diamond                         41.0%  ▁▁▁▁▂●▆▄▅▇█▅ beats 42/211   hard · avg 39%          17%  sharp
AraGen-v3 3C3H score (Inception, …   50.5%  ▁▁·▁▂▂▄█▄●▄▄ beats 44/66    hard · avg 29%         8.2%  at its level
Open PT-LLM Leaderboard average (…   79.3%  ▃▁▁▁▁▁▁▁▂█●▁ beats 253/276  medium · avg 72%       3.6%  at its level · noisy
LongSocialBench (full-context, 4-…   35.6%  ●▂·▄█·▂▂·▄▄▂ beats 0/16     very hard · avg 36%    3.6%  blunt
## 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
Epoch AI                                        3   52% ▇▇▇▇▇▇
Pixtral 12B                                     5   15% ▇▇
LMArena leaderboard dataset                     1   11% ▇
Inception Arabic/Hindi leaderboards             1    8% ▇
CapArena: Benchmarking and Analyzing De…        1    7% ▇
Open Portuguese LLM Leaderboard                 1    4% ▇
LongSocialBench: Do Long-Context LLMs U…        1    4% ▇
prior: pull toward the middle                        1% ▇
weighted average                                   = 50.9
Licences: Epoch AI ([CC BY 4.0](https://creativecommons.org/licenses/by/4.0/)) · Pixtral 12B ([CC BY 4.0](http://creativecommons.org/licenses/by/4.0/)) · LMArena leaderboard dataset ([CC BY 4.0](https://creativecommons.org/licenses/by/4.0/)) · Inception Arabic/Hindi leaderboards ([Apache-2.0](../../licenses/hf-leaderboard-spaces-Apache-2.0.txt)) · CapArena: Benchmarking and Analyzing Detailed Image Captioning in the LLM Era ([CC-BY-4.0](https://creativecommons.org/licenses/by/4.0/)) · Open Portuguese LLM Leaderboard ([Apache-2.0](../../licenses/hf-leaderboard-spaces-Apache-2.0.txt)) · LongSocialBench: Do Long-Context LLMs Understand Online Discussion Threads? ([CC BY 4.0](https://creativecommons.org/licenses/by/4.0/)) ## 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 Llama 3.2 90B Vision Instruct (Q4_K_M, needs ~53.65 GB, from [ollama.com](https://ollama.com/library/llama3.2-vision/tags)) ```sh ollama run llama3.2-vision:90b ``` Other versions: Q8_0 (~106.81 GB) `ollama run llama3.2-vision:90b-instruct-q8_0` Model repository and documentation: [huggingface.co/meta-llama](https://huggingface.co/meta-llama/Llama-3.2-90B-Vision-Instruct) [interactive version](https://whatmodel.sh/model/llama-3-2-90b-vision) · [JSON](https://whatmodel.sh/api?format=json&all=1)