LLM AtlasLLM AtlasSearch models

Leaderboard

reasoning 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
reasoning

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
#1Claude Opus 4.6Anthropic941,000,000
#2Claude Sonnet 4.6Anthropic941,000,000
#3GPT-5.4OpenAI941,000,000
#4Claude 3.7 SonnetAnthropic93200,000
#5Claude Sonnet 4.5Anthropic931,000,000
#6GPT-5.2OpenAI931,000,000
#7GPT-5.4 ProOpenAI931,000,000
#8Gemini 2.5 ProGoogle DeepMind921,048,576
#9Gemini 3.1 ProGoogle DeepMind921,048,576
#10GPT-5.2 ProOpenAI921,000,000
#11Claude Opus 4.1Anthropic91200,000
#12Claude Sonnet 4Anthropic911,000,000
#13Gemini 3.0 ProGoogle DeepMind911,048,576
#14GPT-4oOpenAI91128,000
#15GPT-5OpenAI911,000,000
#16Claude Opus 4Anthropic90200,000
#17Gemini 3.1 FlashGoogle DeepMind901,048,576
#18GPT-5.3-CodexOpenAI901,000,000
#19Gemini 2.5 Pro TTSGoogle DeepMind891,048,576
#20Gemini 2.5 FlashGoogle DeepMind881,048,576
#21Gemini 3.0 FlashGoogle DeepMind881,048,576
#22GPT-5 miniOpenAI881,000,000
#23Command R+ 2026Cohere87128,000
#24GPT-4.1OpenAI871,048,576
#25GPT-5.3 InstantOpenAI871,000,000
#26Mistral Large 25Mistral AI87128,000
#27o3OpenAI87200,000
#28o3-deep-researchOpenAI87200,000
#29Sonar Deep ResearchPerplexity87200,000
#30Claude Haiku 4.5Anthropic86200,000
#31Gemini 2.0 FlashGoogle DeepMind861,048,576
#32Llama 4 MaverickMeta861,048,576
#33o4-miniOpenAI86200,000
#34o4-mini-deep-researchOpenAI86200,000
#35Sonar Reasoning ProPerplexity86200,000
#36Sonar ProPerplexity85200,000
#37Gemini 1.5 ProGoogle DeepMind842,097,152
#38Gemini 2.5 Flash LiveGoogle DeepMind841,048,576
#39Gemini 2.5 Flash Native Audio PreviewGoogle DeepMind841,048,576
#40Gemini 2.5 Flash-LiteGoogle DeepMind841,048,576
#41Claude Opus 3Anthropic83200,000
#42Gemini 3.1 Flash-LiteGoogle DeepMind831,048,576
#43o1OpenAI83200,000
#44Claude Sonnet 3Anthropic82200,000
#45GPT-4.1 miniOpenAI821,048,576
#46Claude Haiku 3.5Anthropic81200,000
#47Gemini 2.0 Flash-LiteGoogle DeepMind811,048,576
#48GPT-5 nanoOpenAI811,000,000
#49gpt-audioOpenAI81128,000
#50gpt-realtimeOpenAI81128,000
#51Nova ProAmazon Web Services81300,000
#52o1-miniOpenAI81128,000
#53SonarPerplexity81128,000
#54Gemini 1.5 FlashGoogle DeepMind801,048,576
#55GPT-4o-miniOpenAI80128,000
#56gpt-oss-120bOpenAI80131,072
#57Llama 3.3 70B InstructMeta80128,000
#58Llama 4 ScoutMeta7810,485,760
#59Step-3.5-FlashStepFun78131,072
#60Claude Haiku 3Anthropic77200,000
#61Claude Haiku 3.5Anthropic77200,000
#62Claude Haiku 4.5Anthropic77200,000
#63CodestralMistral AI77256,000
#64Doubao-Seed-1.6ByteDance / Doubao77128,000
#65Doubao-Seed-1.6-FlashByteDance / Doubao77128,000
#66Doubao-Seed-2.0-CodeByteDance / Doubao77128,000
#67Doubao-Seed-CodeByteDance / Doubao77128,000
#68Llama 3.1 405B InstructMeta77128,000
#69MiniMax-M1MiniMax77204,800
#70MiniMax-Text-01MiniMax77204,800
#71MiniMax-VL-01MiniMax77204,800
#72Voxtral Mini TranscribeMistral AI77131,072
#73Claude Haiku 3Anthropic76200,000
#74Claude Sonnet 3Anthropic76200,000
#75Claude Sonnet 4Anthropic761,000,000
#76CogVideoXZ.AI76128,000
#77CogView 4Z.AI76128,000
#78Command ACohere76256,000
#79Command A ReasoningCohere76128,000
#80Command A TranslateCohere76128,000
#81Command A VisionCohere76128,000
#82Command R+Cohere76128,000
#83Command R7BCohere76128,000
#84DeepSeek-Coder-V2DeepSeek76128,000
#85DeepSeek-Math-V2DeepSeek76128,000
#86DeepSeek-R1DeepSeek76128,000
#87DeepSeek-R1-Distill-Llama-70BDeepSeek76128,000
#88DeepSeek-V2.5DeepSeek76128,000
#89DeepSeek-V3DeepSeek76128,000
#90DeepSeek-V3.1DeepSeek76128,000
#91DeepSeek-V3.1-BaseDeepSeek76128,000
#92DeepSeek-V3.2DeepSeek76128,000
#93DeepSeek-V3.2-ExpDeepSeek76128,000
#94Devstral 2 OpenMistral AI76128,000
#95Devstral Medium 1.0Mistral AI76128,000
#96Devstral Small 2Mistral AI76128,000
#97Embed 4Cohere76128,000
#98ERNIE 3.5 128KBaidu / ERNIE76128,000
#99ERNIE 4.0 Turbo 8KBaidu / ERNIE76128,000
#100ERNIE Functions 8KBaidu / ERNIE76128,000
#101ERNIE Speed 128KBaidu / ERNIE76128,000
#102GLM-4.5Z.AI76128,000
#103GLM-4.5VZ.AI76128,000
#104GLM-4.6Z.AI76128,000
#105GLM-4.6VZ.AI76128,000
#106GLM-4.7Z.AI76128,000
#107GLM-5Z.AI76128,000
#108GLM-ImageZ.AI76128,000
#109GLM-OCRZ.AI76128,000
#110Grok 3xAI76131,072
#111Grok 3 MinixAI76131,072
#112Grok 4xAI76256,000
#113Grok 4 Fast ReasoningxAI76131,072
#114grok-imagexAI76131,072
#115Hunyuan CodeTencent / Hunyuan76128,000
#116Hunyuan LiteTencent / Hunyuan76128,000
#117Hunyuan StandardTencent / Hunyuan76128,000
#118Hunyuan T1Tencent / Hunyuan76256,000
#119Hunyuan T1 VisionTencent / Hunyuan76128,000
#120Hunyuan TurboSTencent / Hunyuan76128,000
#121Hunyuan TurboS LongText 128KTencent / Hunyuan76128,000
#122Jamba 3BAI21 Labs76256,000
#123Jamba LargeAI21 Labs76256,000
#124Jamba Large 1.6AI21 Labs76256,000
#125Jamba MiniAI21 Labs76256,000
#126Jamba Mini 1.6AI21 Labs76256,000
#127Jamba Mini 1.7AI21 Labs76256,000
#128Kimi K2Moonshot AI / Kimi76131,072
#129Kimi K2 ThinkingMoonshot AI / Kimi76256,000
#130Kimi K2 Turbo PreviewMoonshot AI / Kimi76256,000
#131Kimi K2.5Moonshot AI / Kimi76256,000
#132Llama 3.1 70B InstructMeta76128,000
#133Magistral Medium 1.2Mistral AI76128,000
#134Magistral Small 1.2 OpenMistral AI76128,000
#135MiniMax-M2MiniMax76204,800
#136MiniMax-M2.1MiniMax76204,800
#137MiniMax-M2.1-highspeedMiniMax76204,800
#138MiniMax-M2.5MiniMax76204,800
#139MiniMax-M2.5-highspeedMiniMax76204,800
#140Ministral 3 14B OpenMistral AI76128,000
#141Ministral 3 3B OpenMistral AI76128,000
#142Ministral 3 8B OpenMistral AI76128,000
#143Mistral Large 3Mistral AI76128,000
#144Mistral Large 3 OpenMistral AI76128,000
#145Mistral Medium 3.1Mistral AI76128,000
#146Mistral Nemo 12BMistral AI76128,000
#147Mistral Small 3.1Mistral AI76128,000
#148Mistral Small 3.2 OpenMistral AI76128,000
#149Nova LiteAmazon Web Services76300,000
#150Phi-3-vision-128k-instructMicrosoft76128,000
#151Phi-3.5-mini-instructMicrosoft76131,072
#152Phi-3.5-MoE-instructMicrosoft76131,072
#153Phi-3.5-vision-instructMicrosoft76131,072
#154Phi-4-mini-flash-reasoningMicrosoft76131,072
#155Phi-4-mini-instructMicrosoft76131,072
#156Phi-4-multimodal-instructMicrosoft76131,072
#157Phi-4-reasoningMicrosoft76131,072
#158Phi-4-reasoning-plusMicrosoft76131,072
#159Phi-4-reasoning-vision-15BMicrosoft76131,072
#160Pixtral 12BMistral AI76131,072
#161Pixtral LargeMistral AI76131,072
#162Qwen2.5-1.5B-InstructAlibaba Qwen76131,072
#163Qwen2.5-14B-InstructAlibaba Qwen76131,072
#164Qwen2.5-32B-InstructAlibaba Qwen76131,072
#165Qwen2.5-3B-InstructAlibaba Qwen76131,072
#166Qwen2.5-72B-InstructAlibaba Qwen76131,072
#167Qwen2.5-7B-InstructAlibaba Qwen76131,072
#168Qwen2.5-MaxAlibaba Qwen76131,072
#169Qwen2.5-VL-72B-InstructAlibaba Qwen76131,072
#170Qwen2.5-VL-7B-InstructAlibaba Qwen76131,072
#171Qwen3-Coder-NextAlibaba Qwen76131,072
#172Qwen3.5-0.8BAlibaba Qwen76131,072
#173Qwen3.5-122B-A10BAlibaba Qwen76131,072
#174Qwen3.5-27BAlibaba Qwen76131,072
#175Qwen3.5-2BAlibaba Qwen76131,072
#176Qwen3.5-35B-A3BAlibaba Qwen76131,072
#177Qwen3.5-397B-A17BAlibaba Qwen76131,072
#178Qwen3.5-4BAlibaba Qwen76131,072
#179Qwen3.5-9BAlibaba Qwen76131,072
#180Vidu Q1Z.AI76128,000
#181Voxtral Mini OpenMistral AI76131,072
#182Voxtral Small OpenMistral AI76131,072
#183Codestral EmbedMistral AI7532,768
#184ERNIE 4.5 Turbo 32KBaidu / ERNIE7532,768
#185Gemini 1.5 Flash-8BGoogle DeepMind751,048,576
#186Mistral EmbedMistral AI7532,768
#187Mistral ModerationMistral AI7532,768
#188Mistral OCR 2505Mistral AI7532,768
#189warpgrep-v2Morph75128,000
#190Claude Opus 3Anthropic74200,000
#191Claude Opus 4Anthropic74200,000
#192gpt-audio-miniOpenAI74128,000
#193gpt-realtime-miniOpenAI74128,000
#194DeepSeek-OCRDeepSeek7316,384
#195DeepSeek-OCR-2DeepSeek7316,384
#196DeepSeek-VL2-SmallDeepSeek7316,384
#197gpt-oss-20bOpenAI73131,072
#198image-01MiniMax738,192
#199image-01-liveMiniMax738,192
#200Janus-Pro-7BDeepSeek7316,384
#201MiniMax-Speech-02MiniMax738,192
#202morph-v3-fast-applyMorph73128,000
#203music-2.0MiniMax738,192
#204Phi-4Microsoft7316,384
#205Nova MicroAmazon Web Services72128,000
#206Llama 3.2 90B Vision InstructMeta71128,000
#207phi-1Microsoft714,096
#208phi-1_5Microsoft714,096
#209phi-2Microsoft714,096
#210Phi-3-medium-4k-instructMicrosoft714,096
#211Phi-3-mini-4k-instructMicrosoft714,096
#212Phi-tiny-MoE-instructMicrosoft714,096
#213flash-compactMorph70200,000
#214LFM2-24B-A2BLiquid AI7032,768
#215Llama 3.1 8B InstructMeta70128,000
#216Step3-VL-10BStepFun69131,072
#217GPT Image 1OpenAI6832,768
#218MiMo-VL-7BXiaomi68131,072
#219Code Llama 70B InstructMeta678,192
#220GPT-4o mini TranscribeOpenAI67128,000
#221GPT-4o mini TTSOpenAI67128,000
#222GPT-4o TranscribeOpenAI67128,000
#223Llama 3.2 11B Vision InstructMeta67128,000
#224Llama 3.2 1B InstructMeta67128,000
#225Llama 3.2 3B InstructMeta67128,000
#226Meta Llama 3 70B InstructMeta678,192
#227MiMo-Audio-7BXiaomi66131,072
#228chatgpt-image-latestOpenAI6532,768
#229gpt-image-1-miniOpenAI6532,768
#230Llama Guard 4 12BMeta65131,072
#231Step-Audio-R1.1StepFun65131,072
#232LFM2-8B-A1BLiquid AI6432,768
#233Code Llama 34B InstructMeta638,192
#234Meta Llama 3 8B InstructMeta638,192
#235Llama Guard 3 11B VisionMeta62131,072
#236LFM2.5-1.2B-InstructLiquid AI59131,072
#237LFM2.5-1.2B-ThinkingLiquid AI59131,072
#238LFM2-2.6BLiquid AI5532,768
#239pplx-embed-v1-4bPerplexity488,192
#240pplx-embed-v1-0.6bPerplexity468,192
#241Prompt Guard 86MMeta44512
#242FLUX 1.1 ProBlack Forest Labs43512
#243FLUX 1.1 Pro UltraBlack Forest Labs43512
#244FLUX 1 ProBlack Forest Labs41512
#245NextStep-1.1StepFun34512

Why #1: Claude Opus 4.6

Anthropic's most intelligent Claude model for complex agents, coding, and deep reasoning, with 1M token context and 128K output.

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

Why #2: Claude Sonnet 4.6

Anthropic's current Sonnet tier for fast frontier reasoning, coding, and long-context agent work.

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.