whatmodel.sh · find the best AI models you can run

Best AI models for images and documents to run locally on a 64GB graphics card

A leaderboard of the best AI models for images and documents you can run locally on a 64GB graphics card (with 32GB of system RAM). Qwen3.8-Flash-Next is #1 right now, followed by Qwen3.5 397B-A17B and Qwen3.8 27B. Models are ranked on tests that ask questions about images, charts and documents, like MMMU and CharXiv. Each model shows the download that fits and the command to run it.

66 models · scores combine 47 benchmarks · updated how it works · find a leaderboard · benchmark your machine
task[all tasks][agent tasks][coding][chat & writing][math][vision & documents][web search]

Scores use reviewed vision & documents evaluations only. Compare scores within this task. Models without task measurements are excluded.

Models for 32GB RAM · 64GB graphics card → ~64GB fast · local only · reset
+ more
set[all sets][Arena vision (style controlled) · 199][CharXiv descriptive questions · 2]

Filter the model list while keeping task scores. Models without a shared score scale are not yet calibrated; select a set to see its own scores.

Detect your exact chip or graphics card:

curl -G whatmodel.sh/ -d max=0 --data-urlencode "mac=$(sysctl -n hw.memsize machdep.cpu.brand_string)"Mac

curl -G whatmodel.sh/ -d max=0 --data-urlencode "ram=$(awk '/MemT/{print $2}' /proc/meminfo)" --data-urlencode "nv=$(nvidia-smi --query-gpu=name,memory.total --format=csv,noheader 2>/dev/null)"Linux

curl.exe -G whatmodel.sh/ -d max=0 --data-urlencode "ram=$((gcim Win32_ComputerSystem).TotalPhysicalMemory)" --data-urlencode "nv=$(nvidia-smi --query-gpu=name,memory.total --format=csv,noheader 2>$null)"Windows

$ terminal

detecting your OS…

Run curl "whatmodel.sh/?ram=32&gpu=card&vram=64&max=0"to get this list in your terminal (add &format=json for agents).

ram[8][16][24][32][48][64][96][128][192][256]GB
gpu[apple silicon][graphics card][none]
vram[1][2][4][8][12][16][24][32][48][64][80][96][128][160]GB
price[local][$0.05][$0.25][$1][$5][$20][any]max per M

detecting your OS…

Score vs memory to run

#modeltask score↓where↓speed↓RAM↓context↓get itsources
#modeltask score↓where↓speed↓RAM↓context↓get itsources
10Qwen3.8-Flash-Next · 2-bit Alibaba69±4*local~59GB262kllama.cpp › Qwen (Alibaba) vendor-reported benchmark scores for its own models (qwen.ai launch posts, Qwen2.5-VL and Qwen3-VL technical reports, Qwen3.8-Flash-Next card)How Far Is Document Parsing from Solved? PureDocBench: A Source-Traceable Benchmark across Clean, Degraded, and Real-World Settings (arXiv:2605.07492v2)
39Qwen3.5 397B-A17B · 1-bit Alibaba65±2local (slow)~65GB262kllama.cpp › LMArena leaderboard datasetQwen/Qwen3.5-397B-A17B model cardQwen (Alibaba) vendor-reported benchmark scores for its own models (qwen.ai launch posts, Qwen2.5-VL and Qwen3-VL technical reports, Qwen3.8-Flash-Next card)BEAR-Bench: A Bilingual Enterprise and Academic Reasoning Benchmark for Multimodal Models (arXiv:2608.17895v1)Diagram-MMU: A Multi-Modal Benchmark for Scientific Diagrams (arXiv:2608.12262v1)How Far Is Document Parsing from Solved? PureDocBench: A Source-Traceable Benchmark across Clean, Degraded, and Real-World Settings (arXiv:2605.07492v2)
42Qwen3.8 27B · 8-bit Alibaba64±2local~36GB262kOllama › LMArena leaderboard datasetQwen (Alibaba) vendor-reported benchmark scores for its own models (qwen.ai launch posts, Qwen2.5-VL and Qwen3-VL technical reports, Qwen3.8-Flash-Next card)Qwen/Qwen3.8-27B model cardSciDocBench: A Workflow-Centered Benchmark and Data Pipeline for Scientific Document Understanding (arXiv:2609.05141v2)How Far Is Document Parsing from Solved? PureDocBench: A Source-Traceable Benchmark across Clean, Degraded, and Real-World Settings (arXiv:2605.07492v2)
52Gemma 4 31B · 8-bit Google62±2local~36GB262kllama.cpp › LMArena leaderboard datasetGemma Team — Gemma 4 Technical Reportmeta-models/Muse-Glimmer-30B model card (speed and benchmark tables)ChartArena: Benchmarking Chart Parsing across Languages, Scenarios, and Formats (arXiv:2606.01348v3)BEAR-Bench: A Bilingual Enterprise and Academic Reasoning Benchmark for Multimodal Models (arXiv:2608.17895v1)
53Qwen3.6 27B · 8-bit Alibaba62±2local~37GB262kOllama › Qwen/Qwen3.6-27B model cardQwen/Qwen3.8-27B model cardmeta-models/Muse-Glimmer-30B model card (speed and benchmark tables)How Far Is Document Parsing from Solved? PureDocBench: A Source-Traceable Benchmark across Clean, Degraded, and Real-World Settings (arXiv:2605.07492v2)
56Qwen3.5 35B-A3B · 8-bit Alibaba61±2local~47GB262kOllama › Qwen/Qwen3.5-122B-A10B model cardQwen/Qwen3.6-35B-A3B model cardChartArena: Benchmarking Chart Parsing across Languages, Scenarios, and Formats (arXiv:2606.01348v3)Qwen/Qwen3.5-27B model cardHow Far Is Document Parsing from Solved? PureDocBench: A Source-Traceable Benchmark across Clean, Degraded, and Real-World Settings (arXiv:2605.07492v2)
58Qwen3.5 122B-A10B · 4-bit Alibaba61±2local (slow)~79GB262kllama.cpp › Qwen/Qwen3.5-122B-A10B model cardLMArena leaderboard datasetQwen/Qwen3.5-27B model cardHow Far Is Document Parsing from Solved? PureDocBench: A Source-Traceable Benchmark across Clean, Degraded, and Real-World Settings (arXiv:2605.07492v2)
60Qwen3.6 35B-A3B · 8-bit Alibaba61±3local~40GB262kllama.cpp › Qwen/Qwen3.6-27B model cardQwen/Qwen3.6-35B-A3B model cardHow Far Is Document Parsing from Solved? PureDocBench: A Source-Traceable Benchmark across Clean, Degraded, and Real-World Settings (arXiv:2605.07492v2)
61Gemma 4 26B-A4B · 8-bit Google61±2local~30GB262kllama.cpp › LMArena leaderboard datasetGemma Team — Gemma 4 Technical Reportgoogle/diffusiongemma-26B-A4B-it model card
64Qwen3.5 27B · 8-bit Alibaba60±1local~37GB262kOllama › Qwen/Qwen3.5-122B-A10B model cardLMArena leaderboard datasetQwen/Qwen3.6-35B-A3B model cardQwen/Qwen3.5-27B model cardSpaceCast-Bench: Evaluating Predictive Spatial Reasoning in Vision-Language Models (arXiv:2610.12402v1)BEAR-Bench: A Bilingual Enterprise and Academic Reasoning Benchmark for Multimodal Models (arXiv:2608.17895v1)+1
68Muse Glimmer 30B · 8-bit Meta60±4†local~40GB131kOllama › meta-models/Muse-Glimmer-30B model card (speed and benchmark tables)
78Qwen3.5 9B · 8-bit Alibaba57±2local~12GB262kllama.cpp › Qwen/Qwen3.5-9B model cardChartArena: Benchmarking Chart Parsing across Languages, Scenarios, and Formats (arXiv:2606.01348v3)Qwen/Qwen3.5-4B model cardBEAR-Bench: A Bilingual Enterprise and Academic Reasoning Benchmark for Multimodal Models (arXiv:2608.17895v1)SpaceCast-Bench: Evaluating Predictive Spatial Reasoning in Vision-Language Models (arXiv:2610.12402v1)How Far Is Document Parsing from Solved? PureDocBench: A Source-Traceable Benchmark across Clean, Degraded, and Real-World Settings (arXiv:2605.07492v2)
79Inkling Small (266B-A12B) · 2-bit Thinking Machines56±2local (slow)~86GB1Mllama.cpp › LMArena leaderboard datasetthinkingmachines/Inkling-Small model card
80Gemma 4 12B · 8-bit Google56±4†local~15GB262kllama.cpp › Gemma Team — Gemma 4 Technical Report
87Mistral Medium 3.5 (128B) · 2-bit Mistral55±2*local~42GB262kllama.cpp › LMArena leaderboard dataset
96Qwen3-VL 32B Thinking · 8-bit Alibaba54±2†local~39GB256kOllama › Qwen (Alibaba) vendor-reported benchmark scores for its own models (qwen.ai launch posts, Qwen2.5-VL and Qwen3-VL technical reports, Qwen3.8-Flash-Next card)
97Qwen3.5 4B · 8-bit Alibaba53±2local~6.2GB262kllama.cpp › Qwen/Qwen3.5-9B model cardChartArena: Benchmarking Chart Parsing across Languages, Scenarios, and Formats (arXiv:2606.01348v3)Qwen/Qwen3.5-4B model cardBEAR-Bench: A Bilingual Enterprise and Academic Reasoning Benchmark for Multimodal Models (arXiv:2608.17895v1)SpaceCast-Bench: Evaluating Predictive Spatial Reasoning in Vision-Language Models (arXiv:2610.12402v1)How Far Is Document Parsing from Solved? PureDocBench: A Source-Traceable Benchmark across Clean, Degraded, and Real-World Settings (arXiv:2605.07492v2)
101Qwen3-VL 32B Instruct · 8-bit Alibaba52±2†local~39GB256kOllama › Qwen (Alibaba) vendor-reported benchmark scores for its own models (qwen.ai launch posts, Qwen2.5-VL and Qwen3-VL technical reports, Qwen3.8-Flash-Next card)
103Mistral Small 4 (119B-A6.5B) · 2-bit Mistral52±7*†local~40GB262kllama.cpp › mistralai/Mistral-Small-4-119B-2603 model card
106GLM-4.6V · 2-bit Z.ai51±2local~51GB128kllama.cpp › LMArena leaderboard datasetZ.ai GLM-V GitHub repository benchmark image (GLM-4.6V)SciDocBench: A Workflow-Centered Benchmark and Data Pipeline for Scientific Document Understanding (arXiv:2609.05141v2)
107Apriel 1.6 15B Thinker · 8-bit ServiceNow50±4*†local~21GBllama.cpp › ServiceNow-AI/Apriel-1.6-15b-Thinker model card
108GLM-4.5V · 4-bit Z.ai50±2local (slow)~69GB64kllama.cpp › LMArena leaderboard datasetZ.ai GLM-V GitHub repository benchmark image (GLM-4.6V)ServiceNow-AI/Apriel-1.6-15b-Thinker model cardChartArena: Benchmarking Chart Parsing across Languages, Scenarios, and Formats (arXiv:2606.01348v3)
110Qwen3-VL 30B-A3B Thinking · 8-bit Alibaba49±2†local~37GB256kOllama › Qwen (Alibaba) vendor-reported benchmark scores for its own models (qwen.ai launch posts, Qwen2.5-VL and Qwen3-VL technical reports, Qwen3.8-Flash-Next card)
115Qwen3-VL 8B Thinking · 8-bit Alibaba47±2local~12GB256kOllama › Qwen (Alibaba) vendor-reported benchmark scores for its own models (qwen.ai launch posts, Qwen2.5-VL and Qwen3-VL technical reports, Qwen3.8-Flash-Next card)ChartArena: Benchmarking Chart Parsing across Languages, Scenarios, and Formats (arXiv:2606.01348v3)
116Llama 4 Maverick (402B-A17B) · 1-bit Meta47±2local (slow)~65GB1Mllama.cpp › LMArena leaderboard datasetApriel-1.5-15b-Thinker (arXiv:2510.01141v1), Table 4 multimodal benchmarksVMMU: A Vietnamese Multitask Multimodal Understanding and Reasoning Benchmark (arXiv:2508.13680v5)
119Qwen3-VL 30B-A3B Instruct · 8-bit Alibaba46±2†local~37GB256kOllama › Qwen (Alibaba) vendor-reported benchmark scores for its own models (qwen.ai launch posts, Qwen2.5-VL and Qwen3-VL technical reports, Qwen3.8-Flash-Next card)
120Magistral Small 1.2 · 8-bit Mistral46±5*local~32GB131kllama.cpp › Apriel-1.5-15b-Thinker (arXiv:2510.01141v1), Table 4 multimodal benchmarks
124Llama 4 Scout (109B-A17B) · 2-bit Meta45±2*local~36GB328kllama.cpp › LMArena leaderboard datasetVMMU: A Vietnamese Multitask Multimodal Understanding and Reasoning Benchmark (arXiv:2508.13680v5)
125Mistral Small 3.2 24B · 8-bit Mistral44±2local~32GB128kOllama › LMArena leaderboard datasetVMMU: A Vietnamese Multitask Multimodal Understanding and Reasoning Benchmark (arXiv:2508.13680v5)mistralai/Mistral-Small-3.2-24B-Instruct-2506 model cardmistralai/Mistral-Small-4-119B-2603 model card
126Qwen3-VL 4B Thinking · 8-bit Alibaba44±2local~6.9GB256kOllama › Qwen (Alibaba) vendor-reported benchmark scores for its own models (qwen.ai launch posts, Qwen2.5-VL and Qwen3-VL technical reports, Qwen3.8-Flash-Next card)ChartArena: Benchmarking Chart Parsing across Languages, Scenarios, and Formats (arXiv:2606.01348v3)BEAR-Bench: A Bilingual Enterprise and Academic Reasoning Benchmark for Multimodal Models (arXiv:2608.17895v1)
128Qwen2.5-VL 72B Instruct · 4-bit Alibaba44±1local~51GB33kllama.cpp › Qwen (Alibaba) vendor-reported benchmark scores for its own models (qwen.ai launch posts, Qwen2.5-VL and Qwen3-VL technical reports, Qwen3.8-Flash-Next card)LMArena leaderboard datasetQwen2.5-VL-32B-Instruct model cardChartArena: Benchmarking Chart Parsing across Languages, Scenarios, and Formats (arXiv:2606.01348v3)VMMU: A Vietnamese Multitask Multimodal Understanding and Reasoning Benchmark (arXiv:2508.13680v5)Kaleidoscope: In-language Exams for Massively Multilingual Vision Evaluation (arXiv:2504.07072v2)+2
129Qwen3-VL 8B Instruct · 8-bit Alibaba44±2local~12GB256kOllama › Qwen (Alibaba) vendor-reported benchmark scores for its own models (qwen.ai launch posts, Qwen2.5-VL and Qwen3-VL technical reports, Qwen3.8-Flash-Next card)ChartArena: Benchmarking Chart Parsing across Languages, Scenarios, and Formats (arXiv:2606.01348v3)BEAR-Bench: A Bilingual Enterprise and Academic Reasoning Benchmark for Multimodal Models (arXiv:2608.17895v1)How Far Is Document Parsing from Solved? PureDocBench: A Source-Traceable Benchmark across Clean, Degraded, and Real-World Settings (arXiv:2605.07492v2)
131Mistral Small 3.1 24B (2503) · 8-bit Mistral44±2local~32GB128kOllama › LMArena leaderboard datasetMistral AI vendor-reported benchmark scores for its own models (Hugging Face model cards)mistralai/Mistral-Small-3.2-24B-Instruct-2506 model card
133Gemma 3 27B · 8-bit Google42±2local~37GB131kOllama › LMArena leaderboard datasetGemma Team — Gemma 4 Technical ReportGoogle DeepMind Gemma 3 270M IT model cardVisEditBench: Can Vision-Language Models Edit Visualization Code from Multimodal Feedback? (arXiv:2608.10408v1)VMMU: A Vietnamese Multitask Multimodal Understanding and Reasoning Benchmark (arXiv:2508.13680v5)Gemma Team — Gemma 3 Technical Report
135Molmo 2 8B · 4-bit Ai242±2local~38GBhow to run › LMArena leaderboard datasetMolmo2: Open Weights and Data for Vision-Language Models with Video Understanding and Grounding (Ai2, arXiv 2601.10611v4)
136Qwen2.5-VL 32B Instruct · 8-bit Alibaba42±2local~44GB33kOllama › LMArena leaderboard datasetQwen2.5-VL-32B-Instruct model cardApriel-1.5-15b-Thinker (arXiv:2510.01141v1), Table 4 multimodal benchmarksChartArena: Benchmarking Chart Parsing across Languages, Scenarios, and Formats (arXiv:2606.01348v3)Kaleidoscope: In-language Exams for Massively Multilingual Vision Evaluation (arXiv:2504.07072v2)VMMU: A Vietnamese Multitask Multimodal Understanding and Reasoning Benchmark (arXiv:2508.13680v5)
138Gemma 4 E4B · 8-bit Google40±4†local~9.9GB131kllama.cpp › Gemma Team — Gemma 4 Technical Report
139Pixtral Large (2411) · 2-bit Mistral40±2local~51GBllama.cpp › LMArena leaderboard datasetMistral AI vendor-reported benchmark scores for its own models (Hugging Face model cards)
140Qwen3-VL 4B Instruct · 8-bit Alibaba40±2local~6.9GB256kOllama › Qwen (Alibaba) vendor-reported benchmark scores for its own models (qwen.ai launch posts, Qwen2.5-VL and Qwen3-VL technical reports, Qwen3.8-Flash-Next card)ChartArena: Benchmarking Chart Parsing across Languages, Scenarios, and Formats (arXiv:2606.01348v3)How Far Is Document Parsing from Solved? PureDocBench: A Source-Traceable Benchmark across Clean, Degraded, and Real-World Settings (arXiv:2605.07492v2)
141Qwen3.5 2B · 8-bit Alibaba39±2local~4.9GB262kOllama › Qwen/Qwen3.5-2B model cardCohereLabs/North-Micro-Vision-Instruct model card (benchmark table)ChartArena: Benchmarking Chart Parsing across Languages, Scenarios, and Formats (arXiv:2606.01348v3)SpaceCast-Bench: Evaluating Predictive Spatial Reasoning in Vision-Language Models (arXiv:2610.12402v1)How Far Is Document Parsing from Solved? PureDocBench: A Source-Traceable Benchmark across Clean, Degraded, and Real-World Settings (arXiv:2605.07492v2)
143Qwen2-VL 72B Instruct · 4-bit Alibaba38±1local~53GBllama.cpp › Qwen (Alibaba) vendor-reported benchmark scores for its own models (qwen.ai launch posts, Qwen2.5-VL and Qwen3-VL technical reports, Qwen3.8-Flash-Next card)LMArena leaderboard datasetPixtral 12B (Mistral AI, arXiv 2410.07073v2)Qwen2.5-VL-32B-Instruct model cardCapArena: Benchmarking and Analyzing Detailed Image Captioning in the LLM Era (arXiv:2503.12329v1)Ai2 Molmo-72B-0924 model card: Human Preference Elo Rating column
148Qwen2.5-VL 7B Instruct · 8-bit Alibaba35±2local~11GBOllama › Qwen (Alibaba) vendor-reported benchmark scores for its own models (qwen.ai launch posts, Qwen2.5-VL and Qwen3-VL technical reports, Qwen3.8-Flash-Next card)ChartArena: Benchmarking Chart Parsing across Languages, Scenarios, and Formats (arXiv:2606.01348v3)
150Qwen3-VL 2B Thinking · 8-bit Alibaba34±2local~4.2GB256kOllama › Qwen (Alibaba) vendor-reported benchmark scores for its own models (qwen.ai launch posts, Qwen2.5-VL and Qwen3-VL technical reports, Qwen3.8-Flash-Next card)ChartArena: Benchmarking Chart Parsing across Languages, Scenarios, and Formats (arXiv:2606.01348v3)
155Pixtral 12B (2409) · 16-bit Mistral30±2local~28GBhow to run › LMArena leaderboard datasetPixtral 12B (Mistral AI, arXiv 2410.07073v2)VisEditBench: Can Vision-Language Models Edit Visualization Code from Multimodal Feedback? (arXiv:2608.10408v1)Ai2 Molmo-72B-0924 model card: Human Preference Elo Rating column
156Gemma 3 12B · 8-bit Google30±4local~17GB131kOllama › Google DeepMind Gemma 3 270M IT model cardVisEditBench: Can Vision-Language Models Edit Visualization Code from Multimodal Feedback? (arXiv:2608.10408v1)Gemma Team — Gemma 3 Technical Report
157Llama 3.2 90B Vision Instruct · 4-bit Meta29±2local~58GB128kOllama › LMArena leaderboard datasetPixtral 12B (Mistral AI, arXiv 2410.07073v2)CapArena: Benchmarking and Analyzing Detailed Image Captioning in the LLM Era (arXiv:2503.12329v1)
158InternVL2-26B · 16-bit Shanghai AI Laboratory (OpenGVLab)29±2local~55GBhow to run › LMArena leaderboard datasetCharXiv leaderboard (princeton-nlp/CharXiv README)OpenGVLab InternVL2 model cards (InternVL2-4B and InternVL2-26B), Evaluation tablesCapArena: Benchmarking and Analyzing Detailed Image Captioning in the LLM Era (arXiv:2503.12329v1)PhysBench: Benchmarking and Enhancing Vision-Language Models for Physical World Understanding (arXiv:2501.16411v2)
161Qwen2-VL 7B Instruct · 8-bit Alibaba29±1local~11GBllama.cpp › Qwen (Alibaba) vendor-reported benchmark scores for its own models (qwen.ai launch posts, Qwen2.5-VL and Qwen3-VL technical reports, Qwen3.8-Flash-Next card)LMArena leaderboard datasetPixtral 12B (Mistral AI, arXiv 2410.07073v2)CapArena: Benchmarking and Analyzing Detailed Image Captioning in the LLM Era (arXiv:2503.12329v1)Ai2 Molmo-72B-0924 model card: Human Preference Elo Rating column
163Ministral 3 3B · 8-bit Mistral27±4local~6.9GB262kOllama › CohereLabs/North-Micro-Vision-Instruct model card (benchmark table)
164Qwen3-VL 2B Instruct · 8-bit Alibaba27±1local~4.2GB256kOllama › Qwen (Alibaba) vendor-reported benchmark scores for its own models (qwen.ai launch posts, Qwen2.5-VL and Qwen3-VL technical reports, Qwen3.8-Flash-Next card)CohereLabs/North-Micro-Vision-Instruct model card (benchmark table)ChartArena: Benchmarking Chart Parsing across Languages, Scenarios, and Formats (arXiv:2606.01348v3)BEAR-Bench: A Bilingual Enterprise and Academic Reasoning Benchmark for Multimodal Models (arXiv:2608.17895v1)How Far Is Document Parsing from Solved? PureDocBench: A Source-Traceable Benchmark across Clean, Degraded, and Real-World Settings (arXiv:2605.07492v2)
165Gemma 4 E2B · 8-bit Google27±3local~6.8GB131kllama.cpp › CohereLabs/North-Micro-Vision-Instruct model card (benchmark table)Gemma Team — Gemma 4 Technical Report
166Qwen2.5-VL 3B Instruct · 8-bit Alibaba27±2local~6.3GBOllama › Qwen (Alibaba) vendor-reported benchmark scores for its own models (qwen.ai launch posts, Qwen2.5-VL and Qwen3-VL technical reports, Qwen3.8-Flash-Next card)ChartArena: Benchmarking Chart Parsing across Languages, Scenarios, and Formats (arXiv:2606.01348v3)
167Qwen3.5 0.8B · 8-bit Alibaba26±2local~3.1GB262kOllama › Qwen/Qwen3.5-2B model cardChartArena: Benchmarking Chart Parsing across Languages, Scenarios, and Formats (arXiv:2606.01348v3)BEAR-Bench: A Bilingual Enterprise and Academic Reasoning Benchmark for Multimodal Models (arXiv:2608.17895v1)How Far Is Document Parsing from Solved? PureDocBench: A Source-Traceable Benchmark across Clean, Degraded, and Real-World Settings (arXiv:2605.07492v2)
169Aya Vision 32B · 4-bit Cohere25±2local (slow)~71GB16khow to run › LMArena leaderboard datasetKaleidoscope: In-language Exams for Massively Multilingual Vision Evaluation (arXiv:2504.07072v2)VMMU: A Vietnamese Multitask Multimodal Understanding and Reasoning Benchmark (arXiv:2508.13680v5)
171Molmo 7B-D (0924) · 4-bit Ai224±2local~35GBhow to run › LMArena leaderboard datasetMolmo2: Open Weights and Data for Vision-Language Models with Video Understanding and Grounding (Ai2, arXiv 2601.10611v4)Pixtral 12B (Mistral AI, arXiv 2410.07073v2)Kaleidoscope: In-language Exams for Massively Multilingual Vision Evaluation (arXiv:2504.07072v2)Ai2 Molmo-72B-0924 model card: Human Preference Elo Rating column
173Llama 3.2 11B Vision Instruct · 8-bit Meta22±2local~16GB128kOllama › LMArena leaderboard datasetPixtral 12B (Mistral AI, arXiv 2410.07073v2)
175CogVLM2 Llama3 Chat 19B · 16-bit Zhipu AI (THUDM)21±2local~43GBhow to run › LMArena leaderboard datasetCapArena: Benchmarking and Analyzing Detailed Image Captioning in the LLM Era (arXiv:2503.12329v1)Zhipu AI (THUDM) CogVLM2 model card, Benchmark table (own models)
176MiniCPM-V 2.6 · 8-bit OpenBMB21±2local~12GBOllama › LMArena leaderboard datasetCharXiv leaderboard (princeton-nlp/CharXiv README)CapArena: Benchmarking and Analyzing Detailed Image Captioning in the LLM Era (arXiv:2503.12329v1)PhysBench: Benchmarking and Enhancing Vision-Language Models for Physical World Understanding (arXiv:2501.16411v2)
177InternVL2-4B · 16-bit Shanghai AI Laboratory (OpenGVLab)20±2local~10GBhow to run › LMArena leaderboard datasetOpenGVLab InternVL2 model cards (InternVL2-4B and InternVL2-26B), Evaluation tablesMicrosoft Phi-3.5-vision-instruct model card: Benchmarks tablePhysBench: Benchmarking and Enhancing Vision-Language Models for Physical World Understanding (arXiv:2501.16411v2)
178LLaVA-v1.6-34B · 8-bit LLaVA19±2local~46GB4kOllama › LMArena leaderboard datasetCharXiv: Charting Gaps in Realistic Chart Understanding in Multimodal LLMs (arXiv:2406.18521v1)CharXiv leaderboard (princeton-nlp/CharXiv README)CapArena: Benchmarking and Analyzing Detailed Image Captioning in the LLM Era (arXiv:2503.12329v1)LLaVA repository model zoo (haotian-liu/LLaVA docs/MODEL_ZOO.md)
179Phi-3.5-vision-instruct · 16-bit Microsoft16±2local~10GBhow to run › LMArena leaderboard datasetCohereLabs/North-Micro-Vision-Instruct model card (benchmark table)Microsoft Phi-3.5-vision-instruct model card: Benchmarks tableAi2 Molmo-72B-0924 model card: Human Preference Elo Rating columnPhysBench: Benchmarking and Enhancing Vision-Language Models for Physical World Understanding (arXiv:2501.16411v2)
180Qwen2-VL 2B Instruct · 4-bit Alibaba14±2†local~3.3GB33kllama.cpp › Qwen (Alibaba) vendor-reported benchmark scores for its own models (qwen.ai launch posts, Qwen2.5-VL and Qwen3-VL technical reports, Qwen3.8-Flash-Next card)
181Gemma 3 4B · 8-bit Google14±4local~7.4GB131kOllama › Google DeepMind Gemma 3 270M IT model cardVisEditBench: Can Vision-Language Models Edit Visualization Code from Multimodal Feedback? (arXiv:2608.10408v1)Gemma Team — Gemma 3 Technical ReportVMMU: A Vietnamese Multitask Multimodal Understanding and Reasoning Benchmark (arXiv:2508.13680v5)
182Phi-3-vision-128k-instruct · 16-bit Microsoft12±2local~10GBhow to run › LMArena leaderboard datasetCharXiv leaderboard (princeton-nlp/CharXiv README)Phi-3 Technical Report (Microsoft, arXiv 2404.14219v3)PhysBench: Benchmarking and Enhancing Vision-Language Models for Physical World Understanding (arXiv:2501.16411v2)
—Command A+ (05-2026) · iq2_m Coherenot yet calibratedlocal (slow)~71GB128kllama.cpp › Cohere vendor-reported benchmark scores for its own models (cohere.com blog)
—EXAONE 4.5 33B · gguf-q8_0 LG AI Researchnot scored yetlocal~45GB262kllama.cpp ›

+ 118 more don't fit your machine or budget · show them

+ 1 models with fewer than 3 results left off, since one or two tests say little about overall ability · show them

184 of 202 catalogue models listed here · sources and licences

* new or thinly tested: ranked on too few benchmark results · † only the maker’s own numbers so far · ~RAM is estimated; full context may need more memory