LLM AtlasLLM AtlasSearch models

Leaderboard

vision rankings

Updated weekly

Each leaderboard uses transparent weighted scoring, current model context, and supporting analysis to help teams interpret the results with confidence. Only full-profile entries appear in rankings; broader catalog records remain available elsewhere on the site when only source-backed metadata is currently available.

Full-profile entries
245

Models with complete enough metadata and scoring coverage to be meaningfully ranked in this category.

Ranking basis
vision

Scores combine benchmark evidence, product metadata, and cost/context signals when those fields are published.

Catalog caveat
25

Tracked models without full scoring remain in the directory and provider pages, but are not relied on for analytical ranking claims.

RankModelProviderScoreContext
#1Gemini 3.1 ProGoogle DeepMind931,048,576
#2GPT-4oOpenAI93128,000
#3GPT-5.4OpenAI931,000,000
#4Gemini 2.5 ProGoogle DeepMind921,048,576
#5GPT-5.4 ProOpenAI921,000,000
#6Gemini 2.5 Pro TTSGoogle DeepMind911,048,576
#7Gemini 3.0 ProGoogle DeepMind911,048,576
#8Gemini 3.1 FlashGoogle DeepMind911,048,576
#9GPT-5.2OpenAI911,000,000
#10Claude Opus 4.6Anthropic901,000,000
#11Claude Sonnet 4.5Anthropic901,000,000
#12Claude Sonnet 4.6Anthropic901,000,000
#13GPT-5.2 ProOpenAI901,000,000
#14Gemini 2.5 FlashGoogle DeepMind891,048,576
#15Gemini 3.0 FlashGoogle DeepMind891,048,576
#16Claude Sonnet 4Anthropic881,000,000
#17GPT-5OpenAI881,000,000
#18Claude Opus 4.1Anthropic87200,000
#19Gemini 2.0 FlashGoogle DeepMind871,048,576
#20Claude 3.7 SonnetAnthropic86200,000
#21Claude Opus 4Anthropic86200,000
#22Gemini 2.5 Flash LiveGoogle DeepMind851,048,576
#23Gemini 2.5 Flash Native Audio PreviewGoogle DeepMind851,048,576
#24Gemini 2.5 Flash-LiteGoogle DeepMind851,048,576
#25Gemini 3.1 Flash-LiteGoogle DeepMind841,048,576
#26GPT-4.1OpenAI841,048,576
#27Mistral Large 25Mistral AI84128,000
#28GPT Image 1OpenAI8332,768
#29GPT-5 miniOpenAI831,000,000
#30Llama 4 MaverickMeta831,048,576
#31Gemini 1.5 ProGoogle DeepMind822,097,152
#32GPT-5.3 InstantOpenAI821,000,000
#33Claude Haiku 4.5Anthropic80200,000
#34Gemini 2.0 Flash-LiteGoogle DeepMind801,048,576
#35chatgpt-image-latestOpenAI7932,768
#36GPT-5.3-CodexOpenAI791,000,000
#37gpt-image-1-miniOpenAI7932,768
#38Nova ProAmazon Web Services78300,000
#39Claude Opus 3Anthropic77200,000
#40Gemini 1.5 FlashGoogle DeepMind771,048,576
#41GPT-4.1 miniOpenAI761,048,576
#42GPT-5 nanoOpenAI751,000,000
#43o3OpenAI75200,000
#44Claude Haiku 3.5Anthropic74200,000
#45Claude Sonnet 3Anthropic74200,000
#46FLUX 1.1 ProBlack Forest Labs74512
#47FLUX 1.1 Pro UltraBlack Forest Labs74512
#48Llama 3.2 90B Vision InstructMeta74128,000
#49o3-deep-researchOpenAI74200,000
#50o4-mini-deep-researchOpenAI74200,000
#51GPT-4o-miniOpenAI73128,000
#52Command R+ 2026Cohere72128,000
#53o4-miniOpenAI72200,000
#54Step3-VL-10BStepFun72131,072
#55FLUX 1 ProBlack Forest Labs71512
#56Llama 4 ScoutMeta7110,485,760
#57MiMo-VL-7BXiaomi71131,072
#58Gemini 1.5 Flash-8BGoogle DeepMind701,048,576
#59Nova LiteAmazon Web Services70300,000
#60Llama 3.2 11B Vision InstructMeta68128,000
#61Claude Haiku 3Anthropic67200,000
#62Claude Haiku 3.5Anthropic66200,000
#63Claude Haiku 4.5Anthropic66200,000
#64CodestralMistral AI66256,000
#65Codestral EmbedMistral AI6632,768
#66CogVideoXZ.AI66128,000
#67CogView 4Z.AI66128,000
#68Devstral Medium 1.0Mistral AI66128,000
#69Doubao-Seed-1.6ByteDance / Doubao66128,000
#70Doubao-Seed-1.6-FlashByteDance / Doubao66128,000
#71Doubao-Seed-2.0-CodeByteDance / Doubao66128,000
#72Doubao-Seed-CodeByteDance / Doubao66128,000
#73ERNIE 3.5 128KBaidu / ERNIE66128,000
#74ERNIE 4.0 Turbo 8KBaidu / ERNIE66128,000
#75ERNIE Functions 8KBaidu / ERNIE66128,000
#76ERNIE Speed 128KBaidu / ERNIE66128,000
#77GLM-4.5Z.AI66128,000
#78GLM-4.5VZ.AI66128,000
#79GLM-4.6Z.AI66128,000
#80GLM-4.6VZ.AI66128,000
#81GLM-4.7Z.AI66128,000
#82GLM-5Z.AI66128,000
#83GLM-ImageZ.AI66128,000
#84GLM-OCRZ.AI66128,000
#85Hunyuan CodeTencent / Hunyuan66128,000
#86Hunyuan LiteTencent / Hunyuan66128,000
#87Hunyuan StandardTencent / Hunyuan66128,000
#88Hunyuan T1Tencent / Hunyuan66256,000
#89Hunyuan T1 VisionTencent / Hunyuan66128,000
#90Hunyuan TurboSTencent / Hunyuan66128,000
#91Hunyuan TurboS LongText 128KTencent / Hunyuan66128,000
#92Kimi K2Moonshot AI / Kimi66131,072
#93Kimi K2 ThinkingMoonshot AI / Kimi66256,000
#94Kimi K2 Turbo PreviewMoonshot AI / Kimi66256,000
#95Kimi K2.5Moonshot AI / Kimi66256,000
#96Llama 3.3 70B InstructMeta66128,000
#97Magistral Medium 1.2Mistral AI66128,000
#98MiniMax-M1MiniMax66204,800
#99MiniMax-M2MiniMax66204,800
#100MiniMax-M2.1MiniMax66204,800
#101MiniMax-M2.1-highspeedMiniMax66204,800
#102MiniMax-M2.5MiniMax66204,800
#103MiniMax-M2.5-highspeedMiniMax66204,800
#104MiniMax-Text-01MiniMax66204,800
#105MiniMax-VL-01MiniMax66204,800
#106Mistral EmbedMistral AI6632,768
#107Mistral Large 3Mistral AI66128,000
#108Mistral Medium 3.1Mistral AI66128,000
#109Mistral ModerationMistral AI6632,768
#110Mistral Small 3.1Mistral AI66128,000
#111Mistral Small 3.2 OpenMistral AI66128,000
#112Pixtral 12BMistral AI66131,072
#113Pixtral LargeMistral AI66131,072
#114Vidu Q1Z.AI66128,000
#115Voxtral Mini OpenMistral AI66131,072
#116Voxtral Mini TranscribeMistral AI66131,072
#117Voxtral Small OpenMistral AI66131,072
#118Claude Haiku 3Anthropic65200,000
#119Claude Sonnet 3Anthropic65200,000
#120Claude Sonnet 4Anthropic651,000,000
#121Command ACohere65256,000
#122Command A ReasoningCohere65128,000
#123Command A TranslateCohere65128,000
#124Command A VisionCohere65128,000
#125Command R+Cohere65128,000
#126Command R7BCohere65128,000
#127DeepSeek-Coder-V2DeepSeek65128,000
#128DeepSeek-Math-V2DeepSeek65128,000
#129DeepSeek-R1DeepSeek65128,000
#130DeepSeek-R1-Distill-Llama-70BDeepSeek65128,000
#131DeepSeek-V2.5DeepSeek65128,000
#132DeepSeek-V3DeepSeek65128,000
#133DeepSeek-V3.1DeepSeek65128,000
#134DeepSeek-V3.1-BaseDeepSeek65128,000
#135DeepSeek-V3.2DeepSeek65128,000
#136DeepSeek-V3.2-ExpDeepSeek65128,000
#137Devstral 2 OpenMistral AI65128,000
#138Devstral Small 2Mistral AI65128,000
#139Embed 4Cohere65128,000
#140ERNIE 4.5 Turbo 32KBaidu / ERNIE6532,768
#141Grok 3xAI65131,072
#142Grok 3 MinixAI65131,072
#143Grok 4xAI65256,000
#144Grok 4 Fast ReasoningxAI65131,072
#145grok-imagexAI65131,072
#146Jamba 3BAI21 Labs65256,000
#147Jamba LargeAI21 Labs65256,000
#148Jamba Large 1.6AI21 Labs65256,000
#149Jamba MiniAI21 Labs65256,000
#150Jamba Mini 1.6AI21 Labs65256,000
#151Jamba Mini 1.7AI21 Labs65256,000
#152Magistral Small 1.2 OpenMistral AI65128,000
#153Ministral 3 14B OpenMistral AI65128,000
#154Ministral 3 3B OpenMistral AI65128,000
#155Ministral 3 8B OpenMistral AI65128,000
#156Mistral Large 3 OpenMistral AI65128,000
#157Mistral Nemo 12BMistral AI65128,000
#158Mistral OCR 2505Mistral AI6532,768
#159Phi-3-vision-128k-instructMicrosoft65128,000
#160Phi-3.5-mini-instructMicrosoft65131,072
#161Phi-3.5-MoE-instructMicrosoft65131,072
#162Phi-3.5-vision-instructMicrosoft65131,072
#163Phi-4-mini-flash-reasoningMicrosoft65131,072
#164Phi-4-mini-instructMicrosoft65131,072
#165Phi-4-multimodal-instructMicrosoft65131,072
#166Phi-4-reasoningMicrosoft65131,072
#167Phi-4-reasoning-plusMicrosoft65131,072
#168Phi-4-reasoning-vision-15BMicrosoft65131,072
#169Qwen2.5-1.5B-InstructAlibaba Qwen65131,072
#170Qwen2.5-14B-InstructAlibaba Qwen65131,072
#171Qwen2.5-32B-InstructAlibaba Qwen65131,072
#172Qwen2.5-3B-InstructAlibaba Qwen65131,072
#173Qwen2.5-72B-InstructAlibaba Qwen65131,072
#174Qwen2.5-7B-InstructAlibaba Qwen65131,072
#175Qwen2.5-MaxAlibaba Qwen65131,072
#176Qwen2.5-VL-72B-InstructAlibaba Qwen65131,072
#177Qwen2.5-VL-7B-InstructAlibaba Qwen65131,072
#178Qwen3-Coder-NextAlibaba Qwen65131,072
#179Qwen3.5-0.8BAlibaba Qwen65131,072
#180Qwen3.5-122B-A10BAlibaba Qwen65131,072
#181Qwen3.5-27BAlibaba Qwen65131,072
#182Qwen3.5-2BAlibaba Qwen65131,072
#183Qwen3.5-35B-A3BAlibaba Qwen65131,072
#184Qwen3.5-397B-A17BAlibaba Qwen65131,072
#185Qwen3.5-4BAlibaba Qwen65131,072
#186Qwen3.5-9BAlibaba Qwen65131,072
#187DeepSeek-OCRDeepSeek6416,384
#188DeepSeek-OCR-2DeepSeek6416,384
#189DeepSeek-VL2-SmallDeepSeek6416,384
#190image-01MiniMax648,192
#191image-01-liveMiniMax648,192
#192Janus-Pro-7BDeepSeek6416,384
#193MiniMax-Speech-02MiniMax648,192
#194music-2.0MiniMax648,192
#195NextStep-1.1StepFun64512
#196o1OpenAI64200,000
#197Phi-4Microsoft6416,384
#198Claude Opus 3Anthropic63200,000
#199Claude Opus 4Anthropic63200,000
#200phi-1Microsoft634,096
#201phi-1_5Microsoft634,096
#202phi-2Microsoft634,096
#203Phi-3-medium-4k-instructMicrosoft634,096
#204Phi-3-mini-4k-instructMicrosoft634,096
#205Phi-tiny-MoE-instructMicrosoft634,096
#206Llama Guard 4 12BMeta62131,072
#207Llama 3.1 405B InstructMeta61128,000
#208Llama 3.1 70B InstructMeta61128,000
#209gpt-oss-120bOpenAI59131,072
#210MiMo-Audio-7BXiaomi59131,072
#211o1-miniOpenAI59128,000
#212gpt-audioOpenAI58128,000
#213gpt-realtimeOpenAI58128,000
#214Llama Guard 3 11B VisionMeta57131,072
#215Sonar ProPerplexity55200,000
#216Llama 3.1 8B InstructMeta54128,000
#217Sonar Reasoning ProPerplexity54200,000
#218Step-3.5-FlashStepFun54131,072
#219Nova MicroAmazon Web Services53128,000
#220Step-Audio-R1.1StepFun53131,072
#221gpt-audio-miniOpenAI52128,000
#222gpt-oss-20bOpenAI52131,072
#223gpt-realtime-miniOpenAI52128,000
#224SonarPerplexity52128,000
#225Llama 3.2 1B InstructMeta51128,000
#226Llama 3.2 3B InstructMeta51128,000
#227Sonar Deep ResearchPerplexity51200,000
#228Code Llama 70B InstructMeta468,192
#229Meta Llama 3 70B InstructMeta468,192
#230GPT-4o mini TranscribeOpenAI44128,000
#231GPT-4o mini TTSOpenAI44128,000
#232GPT-4o TranscribeOpenAI44128,000
#233Code Llama 34B InstructMeta438,192
#234Meta Llama 3 8B InstructMeta438,192
#235morph-v3-fast-applyMorph43128,000
#236warpgrep-v2Morph42128,000
#237LFM2-24B-A2BLiquid AI4132,768
#238flash-compactMorph39200,000
#239LFM2-8B-A1BLiquid AI3632,768
#240LFM2.5-1.2B-InstructLiquid AI32131,072
#241LFM2.5-1.2B-ThinkingLiquid AI32131,072
#242LFM2-2.6BLiquid AI3132,768
#243pplx-embed-v1-4bPerplexity318,192
#244pplx-embed-v1-0.6bPerplexity308,192
#245Prompt Guard 86MMeta29512

Why #1: Gemini 3.1 Pro

Google's Gemini 3.1 Pro, designed for complex tasks where simple answers aren't enough. Released Feb 2026 with enhanced reasoning and multimodal capabilities.

This model clears the current full-profile threshold for leaderboard methodology.

Why #2: GPT-4o

A broadly capable multimodal model optimized for production chat, agentic workflows, and voice experiences.

This model clears the current full-profile threshold for leaderboard methodology.

Why #3: GPT-5.4

OpenAI's GPT-5.4, the most capable and efficient frontier model for professional work. First general-purpose model with native computer-use capabilities. Combines industry-leading coding from GPT-5.3-Codex with improved agentic workflows.

This model clears the current full-profile threshold for leaderboard methodology.