COMPARE / 2–4 MODELS
模型对比
MODEL SELECTOR
选择对比模型
4 / 4
Claude Fable 5.1 (Adaptive Reasoning, Max Effort, Default Fallback)AnthropicAnthropicGPT-6 Astra (max)OpenAIOpenAIClaude Opus 5 (Adaptive Reasoning, Max Effort)AnthropicAnthropicClaude Fable 5 (Adaptive Reasoning, Max Effort, Opus 4.8 Fallback)AnthropicAnthropicMuse Spark 1.3 (max)MetaMetaGPT-5.6 Sol (max)OpenAIOpenAIQwen3.8 Max (0902)阿里巴巴(通义千问)AlibabaGLM-5.3 (max)智谱 AI(Z.ai)Z AIGrok 4.6 (high)SpaceXAISpaceXAIKimi K3 (max)月之暗面(Kimi)KimiGPT-5.6 Terra (max)OpenAIOpenAIClaude Opus 4.8 (Adaptive Reasoning, Max Effort)AnthropicAnthropicGLM 5.3 Flash智谱 AI(Z.ai)Z AIGemini 3.8 Flash (high)GoogleGoogleClaude Opus 4.7 (Adaptive Reasoning, Max Effort)AnthropicAnthropicQwen3.8 Max阿里巴巴(通义千问)AlibabaQwen3.8 2.4T A95B阿里巴巴(通义千问)AlibabaQwen3.8-Flash-Next阿里巴巴(通义千问)AlibabaMuse Spark 1.2 (xhigh)MetaMetaGemini 3.7 Flash (high)GoogleGoogleDeepSeek V4.1 Flash (Reasoning, Max Effort)深度求索(DeepSeek)DeepSeekGrok 4.5 (high)SpaceXAISpaceXAIGPT-5.4 (xhigh)OpenAIOpenAIGPT-5.5 (xhigh)OpenAIOpenAIClaude Sonnet 5 (Adaptive Reasoning, Max Effort)AnthropicAnthropicGPT-5.6 Luna (max)OpenAIOpenAIDeepSeek V4 Pro 0813 (Reasoning, Max Effort)深度求索(DeepSeek)DeepSeekAgnes 3.0 FlashSapiens AISapiens AIAgnes 2.5 Pro BetaSapiens AISapiens AIDeepSeek V4 Flash Vision (Reasoning, Max Effort)深度求索(DeepSeek)DeepSeekDeepSeek V4 Flash 0731 (Reasoning, Max Effort)深度求索(DeepSeek)DeepSeekGemini 3.6 Flash (high)GoogleGoogleMuse Spark 1.1 (xhigh)MetaMetaGLM-5.2 (max)智谱 AI(Z.ai)Z AIQwen3.8 27B (xhigh)阿里巴巴(通义千问)AlibabaGemini 3.5 Flash (high)GoogleGoogleMotif 3Motif TechnologiesMotif TechnologiesGPT-5.3 Codex (xhigh)OpenAIOpenAIMotif 3 (Beta)Motif TechnologiesMotif TechnologiesClaude Opus 4.6 (Adaptive Reasoning, Max Effort)AnthropicAnthropicMuse SparkMetaMetaDeepSeek V4 Pro 0424 (Reasoning, Max Effort)深度求索(DeepSeek)DeepSeekK2 Horizon 375B A23BInstitute of Foundation ModelsInstitute of Foundation ModelsClaude Sonnet 4.6 (Adaptive Reasoning, Max Effort)AnthropicAnthropicGPT-5.2 (xhigh)OpenAIOpenAIApodex 1.1ApodexApodexGemini 3.1 Pro PreviewGoogleGoogleQwen3.7 Max阿里巴巴(通义千问)AlibabaMiniMax-M3稀宇科技(MiniMax)MiniMaxClaude Opus 4.5 (Reasoning)AnthropicAnthropicMiMo-V2-Pro小米XiaomiGPT-5.2 Codex (xhigh)OpenAIOpenAIQwen3.6 Max Preview阿里巴巴(通义千问)AlibabaNex-N2-ProNex AGINex AGISolar Pro 4UpstageUpstageGemini 3 Pro Preview (high)GoogleGoogleGLM-5 (Reasoning)智谱 AI(Z.ai)Z AIAgnes 2.5 Pro AlphaSapiens AISapiens AIKimi K2.6月之暗面(Kimi)KimiJT-4.1 Flash 236B A21B中国移动China MobileGrok Build 0.1 0616SpaceXAISpaceXAIQuasar 438B (max, based on GLM-5.2)Multiverse ComputingMultiverse ComputingQwen3.6 Plus阿里巴巴(通义千问)AlibabaGPT-5.5 Instant (June 2026)OpenAIOpenAIGLM-5-Turbo智谱 AI(Z.ai)Z AIGLM-5.1 (Reasoning)智谱 AI(Z.ai)Z AIMiMo-V2.5-Pro小米XiaomiGemini 3 Flash Preview (Reasoning)GoogleGoogleKimi K2.7 Code月之暗面(Kimi)KimiInkling SmallThinking MachinesThinking MachinesQwen3.7 Plus阿里巴巴(通义千问)AlibabaHy3腾讯TencentK2 Horizon MoVA 36B A4BInstitute of Foundation ModelsInstitute of Foundation ModelsGrok 4.20 0309 v2 (Reasoning)SpaceXAISpaceXAIInkling (xhigh)Thinking MachinesThinking MachinesGrok 4.3 (high)SpaceXAISpaceXAIGrok 4.20 0309 (Reasoning)SpaceXAISpaceXAIMiMo-V2-Omni-0327小米XiaomiLing-3.0-flash-VL蚂蚁集团(百灵)InclusionAIGPT-5 Codex (high)OpenAIOpenAIDeepSeek V4 Flash 0420 (Reasoning, High Effort)深度求索(DeepSeek)DeepSeekSolar Open2 250BUpstageUpstageGPT-5.1 (high)OpenAIOpenAIGPT-5.4 mini (xhigh)OpenAIOpenAIMiMo-V2-Omni小米XiaomiGPT-5.1 Codex (high)OpenAIOpenAIGLM 5V Turbo (Reasoning)智谱 AI(Z.ai)Z AIKimi K2.5 (Reasoning)月之暗面(Kimi)KimiNemotron 3 Ultra 550B A55B (Reasoning)NVIDIANVIDIAMiniMax-M2.7稀宇科技(MiniMax)MiniMaxLing-3.0-flash-Fin蚂蚁集团(百灵)InclusionAIGPT-5 (high)OpenAIOpenAIQwen3.5 27B (Reasoning)阿里巴巴(通义千问)AlibabaClaude 4.1 Opus (Reasoning)AnthropicAnthropicMiniMax-M2.5稀宇科技(MiniMax)MiniMaxHy3-preview (Reasoning)腾讯TencentA.X-K2SK TelecomSK TelecomGPT-5.5 Instant (May 2026)OpenAIOpenAIGemini 3.5 Flash-LiteGoogleGoogleGrok 4SpaceXAISpaceXAIMiMo-V2-Flash (Feb 2026)小米XiaomiMiMo-V2.5小米XiaomiGLM-4.7 (Reasoning)智谱 AI(Z.ai)Z AIKimi K2 Thinking月之暗面(Kimi)KimiQwen3.6 27B (Reasoning)阿里巴巴(通义千问)Alibabao3-proOpenAIOpenAIG9v3-39A5BAI9StarsAI9StarsKAT Coder Pro V2快手(KAT)KwaiKATDeepSeek V3.2 (Reasoning)深度求索(DeepSeek)DeepSeekQwen3.5 397B A17B (Reasoning)阿里巴巴(通义千问)AlibabaQwen3 Max Thinking阿里巴巴(通义千问)AlibabaGPT-5.4 nano (xhigh)OpenAIOpenAIClaude 4.5 Sonnet (Reasoning)AnthropicAnthropicK2 Horizon 7BInstitute of Foundation ModelsInstitute of Foundation ModelsMiniMax-M2.1稀宇科技(MiniMax)MiniMaxMiMo-V2-Flash (Reasoning)小米XiaomiClaude 4 Opus (Reasoning)AnthropicAnthropicLing 3.0 Flash蚂蚁集团(百灵)InclusionAIGPT-5 mini (high)OpenAIOpenAIQwen3.5 Omni Plus阿里巴巴(通义千问)AlibabaGPT-5.1 Codex mini (high)OpenAIOpenAIGrok 4.1 Fast (Reasoning)SpaceXAISpaceXAIo3OpenAIOpenAIK-EXAONE 2.0 0803LG AI ResearchLG AI ResearchLongCat 2.0美团(LongCat)LongCatStep 3.7 Flash阶跃星辰(StepFun)StepFunQwen3.5 35B A3B (Reasoning)阿里巴巴(通义千问)AlibabaClaude 4 Sonnet (Reasoning)AnthropicAnthropicQwen3.6 35B A3B (Reasoning)阿里巴巴(通义千问)AlibabaJT-35B-Flash中国移动China MobileMiniMax-M2稀宇科技(MiniMax)MiniMaxKAT-Coder-Pro V1快手(KAT)KwaiKATGLM-4.6 (Reasoning)智谱 AI(Z.ai)Z AIMuse Glimmer (high)MetaMetaGrok 4 Fast (Reasoning)SpaceXAISpaceXAIQwen3.5 122B A10B (Reasoning)阿里巴巴(通义千问)AlibabaClaude 3.7 Sonnet (Reasoning)AnthropicAnthropicClaude 4.5 Haiku (Reasoning)AnthropicAnthropicRing-2.6-1T蚂蚁集团(百灵)InclusionAILing-2.6-1T蚂蚁集团(百灵)InclusionAIStep 3.5 Flash 2603阶跃星辰(StepFun)StepFunDoubao Seed Code字节跳动(豆包)ByteDance SeedGemma 4 26B A4B (Reasoning)GoogleGoogleGemini 2.5 ProGoogleGoogleo4-mini (high)OpenAIOpenAIStep 3.5 Flash阶跃星辰(StepFun)StepFunDeepSeek V3.2 Exp (Reasoning)深度求索(DeepSeek)DeepSeekQwen3 Max Thinking (Preview)阿里巴巴(通义千问)AlibabaK2 Horizon 3.7BInstitute of Foundation ModelsInstitute of Foundation ModelsGemini 3.1 Flash-LiteGoogleGoogleQwen3 Max阿里巴巴(通义千问)AlibabaGemini 2.5 Flash Preview (Sep '25) (Reasoning)GoogleGoogleGemma 4 31B (Reasoning)GoogleGoogleDeepSeek V3.1 Terminus (Reasoning)深度求索(DeepSeek)DeepSeekKimi K2 0905月之暗面(Kimi)Kimio1OpenAIOpenAIGemini 2.5 Pro Preview (Mar' 25)GoogleGoogleMistral Medium 3.5MistralMistralGLM-4.7-Flash (Reasoning)智谱 AI(Z.ai)Z AIGranite 4.2 30BIBMIBMGrok 3 mini Reasoning (high)SpaceXAISpaceXAIGemini 2.5 Pro Preview (May' 25)GoogleGoogleDeepSeek V3.2 Speciale深度求索(DeepSeek)DeepSeekK-EXAONE (Reasoning)LG AI ResearchLG AI ResearchERNIE 5.0 Thinking Preview百度BaiduGemma 4 12B (Reasoning)GoogleGoogleNova 2.0 Pro Preview (medium)AmazonAmazonGrok Code Fast 1SpaceXAISpaceXAICommand A+CohereCohereApriel-v1.5-15B-ThinkerServiceNowServiceNowDeepSeek V3.1 (Reasoning)深度求索(DeepSeek)DeepSeekQwen3.5 9B (Reasoning)阿里巴巴(通义千问)AlibabaNova 2.0 Omni (medium)AmazonAmazonNemotron 3.5 LightningNVIDIANVIDIANemotron 3 Super 120B A12B (Reasoning)NVIDIANVIDIAQwen3 VL 235B A22B (Reasoning)阿里巴巴(通义千问)AlibabaApriel-v1.6-15B-ThinkerServiceNowServiceNowNova 2.0 Lite (high)AmazonAmazonEXAONE 4.5 33BLG AI ResearchLG AI ResearchMiniCPM5-2B面壁智能(OpenBMB)OpenBMBQwen3.5 4B (Reasoning)阿里巴巴(通义千问)AlibabaDeepSeek R1 0528 (May '25)深度求索(DeepSeek)DeepSeekGemini 2.5 Flash (Reasoning)GoogleGoogleGPT-5 nano (high)OpenAIOpenAIGLM-4.5 (Reasoning)智谱 AI(Z.ai)Z AIKimi K2月之暗面(Kimi)KimiQwen3 235B A22B 2507 (Reasoning)阿里巴巴(通义千问)AlibabaGPT-4.1OpenAIOpenAIQwen3 Max (Preview)阿里巴巴(通义千问)AlibabaQwen3.5 Omni Flash阿里巴巴(通义千问)Alibabao3-mini (high)OpenAIOpenAIo1-proOpenAIOpenAIgpt-oss-120b (high)OpenAIOpenAIJT-MINI中国移动China MobileGrok 3SpaceXAISpaceXAISeed-OSS-36B-Instruct字节跳动(豆包)ByteDance SeedQwen3 235B A22B 2507 Instruct阿里巴巴(通义千问)AlibabaQwen3 Coder 480B A35B Instruct阿里巴巴(通义千问)AlibabaQwen3 VL 32B (Reasoning)阿里巴巴(通义千问)AlibabaLing 3.0 Tiny蚂蚁集团(百灵)InclusionAIMagistral Medium 1.2MistralMistralSonar Reasoning ProPerplexityPerplexityGranite 4.2 8BIBMIBMHyperNova 60B 2605 (high, based on gpt-oss-120b)Multiverse ComputingMultiverse ComputingMiniMax M1 80k稀宇科技(MiniMax)MiniMaxNemotron Cascade 2 30B A3BNVIDIANVIDIAGemini 2.5 Flash Preview (Reasoning)GoogleGoogleMercury 2InceptionInceptionK2 Think V2Institute of Foundation ModelsInstitute of Foundation ModelsLongCat Flash Lite美团(LongCat)LongCatMistral Small 4 (Reasoning)MistralMistralDeepSeek R1 (Jan '25)深度求索(DeepSeek)DeepSeeko1-previewOpenAIOpenAIHyperCLOVA X SEED Think (32B)NaverNaverGLM-4.6V (Reasoning)智谱 AI(Z.ai)Z AIQwen3 Next 80B A3B (Reasoning)阿里巴巴(通义千问)AlibabaGLM-4.5-Air智谱 AI(Z.ai)Z AIMi:dm K 2.5 ProKorea TelecomKorea TelecomRing-1T蚂蚁集团(百灵)InclusionAITrinity Large ThinkingArcee AIArcee AIG9v3-3BAI9StarsAI9StarsINTELLECT-3Prime IntellectPrime IntellectGPT-5 (ChatGPT)OpenAIOpenAISolar Open 100B (Reasoning)UpstageUpstageGrok 3 Reasoning BetaSpaceXAISpaceXAIGemini 2.5 Flash-Lite Preview (Sep '25) (Reasoning)GoogleGoogleNemotron 3 Nano Omni 30B A3B ReasoningNVIDIANVIDIAGPT-4.1 miniOpenAIOpenAIQwen3 Coder Next阿里巴巴(通义千问)AlibabaMiniMax M1 40k稀宇科技(MiniMax)MiniMaxgpt-oss-20b (high)OpenAIOpenAIQwen3 VL 235B A22B Instruct阿里巴巴(通义千问)AlibabaNorth Mini CodeCohereCohereK2-V2 (high)Institute of Foundation ModelsInstitute of Foundation ModelsQwen3 30B A3B 2507 (Reasoning)阿里巴巴(通义千问)Alibabao1-miniOpenAIOpenAIDeepSeek V3 0324深度求索(DeepSeek)DeepSeekMistral Large 3MistralMistralLing 2.6 Flash蚂蚁集团(百灵)InclusionAIQwen3 Next 80B A3B Instruct阿里巴巴(通义千问)AlibabaTri-21B-think PreviewTrillion LabsTrillion LabsQwen3 Coder 30B A3B Instruct阿里巴巴(通义千问)AlibabaGPT-4.5 (Preview)OpenAIOpenAIMistral Medium 3.1MistralMistralDiffusionGemma 26B A4BGoogleGoogleQwen3 235B A22B (Reasoning)阿里巴巴(通义千问)AlibabaQwQ 32B阿里巴巴(通义千问)AlibabaQwen3 VL 30B A3B (Reasoning)阿里巴巴(通义千问)AlibabaGemini 2.0 Flash Thinking Experimental (Jan '25)GoogleGoogleDevstral 2MistralMistralLlama 4 MaverickMetaMetaMotif-2-12.7B-ReasoningMotif TechnologiesMotif TechnologiesLing-1T蚂蚁集团(百灵)InclusionAINova PremierAmazonAmazonSolar Pro 2 (Preview) (Reasoning)UpstageUpstageGranite 4.2 3BIBMIBMMagistral Medium 1MistralMistralMistral Medium 3MistralMistralLlama Nemotron Super 49B v1.5 (Reasoning)NVIDIANVIDIADevstral MediumMistralMistralTri-21B-ThinkTrillion LabsTrillion LabsGPT-4o (March 2025, chatgpt-4o-latest)OpenAIOpenAIGemini 2.0 Flash (Feb '25)GoogleGoogleClaude 3.5 HaikuAnthropicAnthropicLlama 3.3 Nemotron Super 49B v1 (Reasoning)NVIDIANVIDIAGemma 4 E4B (Reasoning)GoogleGoogleNVIDIA Nemotron 3 Nano 30B A3B (Reasoning)NVIDIANVIDIAQwen3 4B 2507 (Reasoning)阿里巴巴(通义千问)AlibabaMiniCPM5-1B (Reasoning)面壁智能(OpenBMB)OpenBMBSarvam 105B (high)SarvamSarvamGemini 2.0 Pro Experimental (Feb '25)GoogleGoogleDevstral Small (May '25)MistralMistralClaude 3 OpusAnthropicAnthropicSonar ReasoningPerplexityPerplexityMagistral Small 1.2MistralMistralGemini 2.5 Flash-Lite (Reasoning)GoogleGoogleDeepSeek V3 (Dec '24)深度求索(DeepSeek)DeepSeekGPT-4o (Nov '24)OpenAIOpenAINanbeige4.1-3BNanbeigeNanbeigeLFM2.5-2.6BLiquid AILiquid AIQwen3 VL 32B Instruct阿里巴巴(通义千问)AlibabaDeepSeek R1 Distill Qwen 32B深度求索(DeepSeek)DeepSeekDevstral Small 2MistralMistralMagistral Small 1MistralMistralGemini 2.0 Flash (experimental)GoogleGoogleEXAONE 4.0 32B (Reasoning)LG AI ResearchLG AI ResearchQwen3 VL 8B (Reasoning)阿里巴巴(通义千问)AlibabaDeepSeek R1 0528 Qwen3 8B深度求索(DeepSeek)DeepSeekQwen2.5 Max阿里巴巴(通义千问)AlibabaQwen3 VL 30B A3B Instruct阿里巴巴(通义千问)AlibabaHermes 4 - Llama-3.1 70B (Reasoning)Nous ResearchNous ResearchGemini 1.5 Pro (Sep '24)GoogleGoogleDeepSeek R1 Distill Llama 70B深度求索(DeepSeek)DeepSeekClaude 3.5 Sonnet (Oct '24)AnthropicAnthropicDeepSeek R1 Distill Qwen 14B深度求索(DeepSeek)DeepSeekFalcon-H1R-7BTII UAETII UAEGPT-4.1 nanoOpenAIOpenAISolar Pro 3UpstageUpstageLing-flash-2.0蚂蚁集团(百灵)InclusionAIGemma 4 E2B (Reasoning)GoogleGoogleQwen3 Omni 30B A3B (Reasoning)阿里巴巴(通义千问)AlibabaGPT-4o (Aug '24)OpenAIOpenAIQwen2.5 Instruct 72B阿里巴巴(通义千问)AlibabaSonarPerplexityPerplexityStep3 VL 10B阶跃星辰(StepFun)StepFunLlama 3.3 Instruct 70BMetaMetaQwen3 30B A3B (Reasoning)阿里巴巴(通义千问)AlibabaSonar ProPerplexityPerplexityDevstral Small (Jul '25)MistralMistralQwQ 32B-Preview阿里巴巴(通义千问)AlibabaGLM-4.5V (Reasoning)智谱 AI(Z.ai)Z AIMistral Large 2 (Nov '24)MistralMistralLlama 3.1 Nemotron Ultra 253B v1 (Reasoning)NVIDIANVIDIAQwen3 30B A3B 2507 Instruct阿里巴巴(通义千问)AlibabaERNIE 4.5 300B A47B百度BaiduHermes 4 - Llama-3.1 405B (Reasoning)Nous ResearchNous ResearchSolar Pro 2 (Reasoning)UpstageUpstageNVIDIA Nemotron Nano 12B v2 VL (Reasoning)NVIDIANVIDIAGranite 4.1 30BIBMIBMNVIDIA Nemotron Nano 9B V2 (Reasoning)NVIDIANVIDIAGemini 2.0 Flash-Lite (Feb '25)GoogleGoogleNVIDIA Nemotron 3 Nano 4BNVIDIANVIDIAMistral Small 3.1MistralMistralQwen3 32B (Reasoning)阿里巴巴(通义千问)AlibabaGPT-4o (May '24)OpenAIOpenAIGemini 2.0 Flash-Lite (Preview)GoogleGoogleLlama 3.1 Nemotron Nano 4B v1.1 (Reasoning)NVIDIANVIDIAKimi Linear 48B A3B Instruct月之暗面(Kimi)KimiLlama 3.1 Instruct 405BMetaMetaQwen3 VL 8B Instruct阿里巴巴(通义千问)AlibabaQwen3 4B (Reasoning)阿里巴巴(通义千问)AlibabaLFM2.5-8B-A1BLiquid AILiquid AIClaude 3.5 Sonnet (June '24)AnthropicAnthropicLlama 3.1 Tulu3 405BAllen Institute for AIAllen Institute for AIGPT-4o (ChatGPT)OpenAIOpenAIRing-flash-2.0蚂蚁集团(百灵)InclusionAIPixtral LargeMistralMistralOlmo 3.1 32B ThinkAllen Institute for AIAllen Institute for AIGrok 2 (Dec '24)SpaceXAISpaceXAIGemini 1.5 Flash (Sep '24)GoogleGoogleQwen3 VL 4B (Reasoning)阿里巴巴(通义千问)AlibabaGPT-4 TurboOpenAIOpenAIMistral Small 3.2MistralMistralNova ProAmazonAmazonCommand ACohereCohereQwen3.5 2B (Reasoning)阿里巴巴(通义千问)AlibabaLlama 3.1 Nemotron Instruct 70BNVIDIANVIDIALlama 3.1 Instruct 8BMetaMetaGrok BetaSpaceXAISpaceXAIQwen2.5 Instruct 32B阿里巴巴(通义千问)AlibabaMistral Large 2 (Jul '24)MistralMistralQwen3 4B 2507 Instruct阿里巴巴(通义千问)AlibabaQwen2.5 Coder Instruct 32B阿里巴巴(通义千问)AlibabaQwen3 14B (Reasoning)阿里巴巴(通义千问)AlibabaGPT-4OpenAIOpenAIMistral Small 3MistralMistralNova LiteAmazonAmazonGPT-4o miniOpenAIOpenAIDeepSeek-V2.5 (Dec '24)深度求索(DeepSeek)DeepSeekLlama 3.1 Instruct 70BMetaMetaGranite 4.1 8BIBMIBMSarvam 30B (high)SarvamSarvamGemini 2.0 Flash Thinking Experimental (Dec '24)GoogleGoogleDeepSeek-V2.5深度求索(DeepSeek)DeepSeekOlmo 3.1 32B InstructAllen Institute for AIAllen Institute for AIMistral SabaMistralMistralDeepSeek R1 Distill Llama 8B深度求索(DeepSeek)DeepSeekOlmo 3 32B ThinkAllen Institute for AIAllen Institute for AILlama 4 ScoutMetaMetaGemini 1.5 Pro (May '24)GoogleGoogleR1 1776PerplexityPerplexityQwen2.5 Turbo阿里巴巴(通义千问)AlibabaReka Flash (Sep '24)Reka AIReka AILlama 3.2 Instruct 90B (Vision)MetaMetaSolar MiniUpstageUpstageCeleris-1CelerisCelerisGrok-1SpaceXAISpaceXAIQwen2 Instruct 72B阿里巴巴(通义千问)AlibabaPhi-4 Mini InstructMicrosoftMicrosoftGemini 1.5 Flash-8BGoogleGoogleQwen3.5 0.8B (Reasoning)阿里巴巴(通义千问)AlibabaDeepHermes 3 - Mistral 24B Preview (Non-reasoning)Nous ResearchNous ResearchJamba 1.7 LargeAI21 LabsAI21 LabsGranite 4.0 H SmallIBMIBMMinistral 3 14BMistralMistralJamba 1.5 LargeAI21 LabsAI21 LabsQwen3 Omni 30B A3B Instruct阿里巴巴(通义千问)AlibabaHermes 3 - Llama-3.1 70BNous ResearchNous ResearchQwen3 8B (Reasoning)阿里巴巴(通义千问)AlibabaDeepSeek-Coder-V2深度求索(DeepSeek)DeepSeekOLMo 2 32BAllen Institute for AIAllen Institute for AIJamba 1.6 LargeAI21 LabsAI21 LabsLFM2 24B A2BLiquid AILiquid AIGemini 1.5 Flash (May '24)GoogleGooglePhi-4MicrosoftMicrosoftClaude 3 SonnetAnthropicAnthropicNova MicroAmazonAmazonGranite 4.1 3BIBMIBMMistral Small (Sep '24)MistralMistralGemini 1.0 UltraGoogleGooglePhi-3 Mini Instruct 3.8BMicrosoftMicrosoftGemma 3n E4B Instruct Preview (May '25)GoogleGooglePhi-4 Multimodal InstructMicrosoftMicrosoftQwen2.5 Coder Instruct 7B阿里巴巴(通义千问)AlibabaMistral Large (Feb '24)MistralMistralMixtral 8x22B InstructMistralMistralLlama 2 Chat 7BMetaMetaLlama 3.2 Instruct 3BMetaMetaMiniCPM-V 4.6 1.3B面壁智能(OpenBMB)OpenBMBJamba Reasoning 3BAI21 LabsAI21 LabsQwen3 VL 4B Instruct阿里巴巴(通义千问)AlibabaQwen1.5 Chat 110B阿里巴巴(通义千问)AlibabaReka Flash 3Reka AIReka AIOlmo 3 7B ThinkAllen Institute for AIAllen Institute for AIClaude 2.1AnthropicAnthropicClaude 3 HaikuAnthropicAnthropicOLMo 2 7BAllen Institute for AIAllen Institute for AIMolmo 7B-DAllen Institute for AIAllen Institute for AILing-mini-2.0蚂蚁集团(百灵)InclusionAIDeepSeek R1 Distill Qwen 1.5B深度求索(DeepSeek)DeepSeekClaude 2.0AnthropicAnthropicDeepSeek-V2-Chat深度求索(DeepSeek)DeepSeekMistral Small (Feb '24)MistralMistralMistral MediumMistralMistralGPT-3.5 TurboOpenAIOpenAIMinistral 3 8BMistralMistralLlama 3 Instruct 70BMetaMetaArctic InstructSnowflakeSnowflakeQwen Chat 72B阿里巴巴(通义千问)AlibabaLFM 40BLiquid AILiquid AILlama 3.2 Instruct 11B (Vision)MetaMetaPALM-2GoogleGoogleGemini 1.0 ProGoogleGoogleDeepSeek Coder V2 Lite Instruct深度求索(DeepSeek)DeepSeekSarvam M (Reasoning)SarvamSarvamDeepSeek LLM 67B Chat (V1)深度求索(DeepSeek)DeepSeekLlama 2 Chat 70BMetaMetaLlama 2 Chat 13BMetaMetaCommand-R+ (Apr '24)CohereCohereOpenChat 3.5 (1210)OpenChatOpenChatDBRX InstructDatabricksDatabricksExaone 4.0 1.2B (Reasoning)LG AI ResearchLG AI ResearchOlmo 3 7B InstructAllen Institute for AIAllen Institute for AILFM2.5-1.2B-ThinkingLiquid AILiquid AIJamba 1.7 MiniAI21 LabsAI21 LabsLFM2 2.6BLiquid AILiquid AILFM2.5-1.2B-InstructLiquid AILiquid AIJamba 1.5 MiniAI21 LabsAI21 LabsGranite 4.0 H 1BIBMIBMQwen3 1.7B (Reasoning)阿里巴巴(通义千问)AlibabaJamba 1.6 MiniAI21 LabsAI21 LabsMixtral 8x7B InstructMistralMistralGemma 3 270MGoogleGoogleApertus 70B InstructSwiss AI InitiativeSwiss AI InitiativeGranite 4.0 MicroIBMIBMDeepHermes 3 - Llama-3.1 8B Preview (Non-reasoning)Nous ResearchNous ResearchClaude InstantAnthropicAnthropicCommand-R (Mar '24)CohereCohereLlama 65BMetaMetaMistral 7B InstructMistralMistralQwen Chat 14B阿里巴巴(通义千问)AlibabaGranite 4.0 1BIBMIBMMolmo2-8BAllen Institute for AIAllen Institute for AILFM2 8B A1BLiquid AILiquid AIGranite 3.3 8B (Non-reasoning)IBMIBMGemma 3 27B InstructGoogleGoogleMinistral 3 3BMistralMistralApertus 8B InstructSwiss AI InitiativeSwiss AI InitiativeGemma 3 1B InstructGoogleGoogleGemma 3 4B InstructGoogleGoogleGemma 3n E2B InstructGoogleGoogleGemma 3n E4B InstructGoogleGoogleGranite 4.0 350MIBMIBMGranite 4.0 H 350MIBMIBMLFM2 1.2BLiquid AILiquid AILFM2.5-VL-1.6BLiquid AILiquid AILlama 3 Instruct 8BMetaMetaLlama 3.2 Instruct 1BMetaMetaQwen3 0.6B (Reasoning)阿里巴巴(通义千问)AlibabaTiny Aya GlobalCohereCohereGemma 3 12B InstructGoogleGoogleK2 Horizon 0.9BInstitute of Foundation ModelsInstitute of Foundation ModelsCogito v2.1 (Reasoning)Deep CogitoDeep CogitoGemini 3 Deep ThinkGoogleGoogleGPT-3.5 Turbo (0613)OpenAIOpenAIGPT-4o mini Realtime (Dec '24)OpenAIOpenAIGPT-4o Realtime (Dec '24)OpenAIOpenAIGPT-5.4 Pro (xhigh)OpenAIOpenAIGPT-5.5 Pro (xhigh)OpenAIOpenAIMi:dm K 2.5 Pro PreviewKorea TelecomKorea Telecom
当前条件下没有可选模型。
Gemma 4 E4B (Reasoning) / DeepSeek-Coder-V2 / Devstral Small 2 / Command A
PRICE / CAPABILITY FRONTIER
模型斩杀线坐标
3 个所选模型
左上区域代表更低价格和更高能力。红色前沿线由当前主流模型计算;对比状态只显示所选模型点,隐藏其他背景点。价格采用标准输入档位,横轴为对数刻度。
020406080100
AA 智能指数越高越好
Gemma 4 E4B (Reasoning)Google
8.915最优
DeepSeek-Coder-V2DeepSeek
5.984差 32.9%
Devstral Small 2Mistral
8.358差 6.3%
Command ACohere
6.958差 21.9%
GPQA 科学推理越高越好
Gemma 4 E4B (Reasoning)Google
0.576最优
DeepSeek-Coder-V2DeepSeek
—
Devstral Small 2Mistral
0.532差 7.5%
Command ACohere
0.527差 8.4%
τ² 智能体工具任务越高越好
Gemma 4 E4B (Reasoning)Google
0.208差 11.3%
DeepSeek-Coder-V2DeepSeek
—
Devstral Small 2Mistral
0.234最优
Command ACohere
0.152差 35.0%
SciCode 科研编程越高越好
Gemma 4 E4B (Reasoning)Google
—
DeepSeek-Coder-V2DeepSeek
—
Devstral Small 2Mistral
0.324最优
Command ACohere
—
输出速度中位数越高越好
Gemma 4 E4B (Reasoning)Google
44.41差 66.1%
DeepSeek-Coder-V2DeepSeek
—
Devstral Small 2Mistral
130.953最优
Command ACohere
57.404差 56.2%
条长为同一维度内的 0至100 归一化分数;右侧数字为来源原始值。指数使用原始百分制,速度和延迟按本次所选模型相对归一化。
评测数据来自 Artificial Analysis,本站仅作非官方中文呈现;归一化仅用于当前所选模型的视觉比较,不改变原始值。查看指标口径。
FULL SPECIFICATION
完整指标对照
35 项
| 指标 | Gemma 4 E4B (Reasoning) | DeepSeek-Coder-V2 | Devstral Small 2 | Command A |
|---|---|---|---|---|
| AA 智能指数 · 越高越好 | 8.915最优 | 5.984距最优 32.9% | 8.358距最优 6.3% | 6.958距最优 21.9% |
| 智能指数评测成本USD · 越低越好 | —无数据 | —无数据 | 0最优 | —无数据 |
| CritPT 批判性推理 · 越高越好 | 0.006最优 | —无数据 | 0距最优 100.0% | 0距最优 100.0% |
| GDPval 专业任务 · 越高越好 | 0距最优 100.0% | —无数据 | 0.09最优 | —无数据 |
| GPQA 科学推理 · 越高越好 | 0.576最优 | —无数据 | 0.532距最优 7.5% | 0.527距最优 8.4% |
| Humanity’s Last Exam · 越高越好 | 0.038距最优 6.9% | —无数据 | 0.035距最优 13.8% | 0.04最优 |
| IFBench 指令遵循 · 越高越好 | 0.442最优 | —无数据 | 0.312距最优 29.5% | 0.365距最优 17.5% |
| LCR 长上下文推理 · 越高越好 | 0.32最优 | —无数据 | 0.28距最优 12.5% | 0.213距最优 33.3% |
| MMMU-Pro 多模态理解 · 越高越好 | 0.514最优 | —无数据 | 0.446距最优 13.2% | —无数据 |
| Omniscience 综合知识 · 越高越好 | -19.7最优 | —无数据 | -57.7距最优 192.9% | -48.267距最优 145.0% |
| Omniscience 准确率 · 越高越好 | 0.086距最优 47.6% | —无数据 | 0.158距最优 3.6% | 0.164最优 |
| Omniscience 非幻觉率 · 越高越好 | 0.691最优 | —无数据 | 0.127距最优 81.6% | 0.227距最优 67.2% |
| SciCode 科研编程 · 越高越好 | —无数据 | —无数据 | 0.324最优 | —无数据 |
| τ² 智能体工具任务 · 越高越好 | 0.208距最优 11.3% | —无数据 | 0.234最优 | 0.152距最优 35.0% |
| τ-Bench 银行业任务 · 越高越好 | —无数据 | —无数据 | 0.107最优 | —无数据 |
| Terminal-Bench Hard · 越高越好 | 0.083距最优 50.0% | —无数据 | 0.167最优 | 0.008距最优 95.5% |
| 输出速度中位数tokens/s · 越高越好 | 44.41距最优 66.1% | —无数据 | 130.953最优 | 57.404距最优 56.2% |
| 输出速度 P05tokens/s · 越高越好 | 21.396距最优 71.5% | —无数据 | 75.171最优 | 23.002距最优 69.4% |
| 输出速度 P25tokens/s · 越高越好 | 36.763距最优 62.5% | —无数据 | 98.163最优 | 37.109距最优 62.2% |
| 输出速度 P75tokens/s · 越高越好 | 50.246距最优 66.9% | —无数据 | 151.973最优 | 64.2距最优 57.8% |
| 输出速度 P95tokens/s · 越高越好 | 57.009距最优 67.3% | —无数据 | 174.142最优 | 72.751距最优 58.2% |
| 推理耗时中位数秒 · 越低越好 | 45.035最优 | —无数据 | —无数据 | —无数据 |
| 首 token 延迟中位数秒 · 越低越好 | 0.865最优 | —无数据 | 2.293距最优 165.0% | 1.819距最优 110.3% |
| 首 token 延迟 P05秒 · 越低越好 | 0.727最优 | —无数据 | 2.127距最优 192.6% | 1.544距最优 112.5% |
| 首 token 延迟 P25秒 · 越低越好 | 0.789最优 | —无数据 | 2.216距最优 180.9% | 1.7距最优 115.5% |
| 首 token 延迟 P75秒 · 越低越好 | 0.982最优 | —无数据 | 2.732距最优 178.3% | 2.294距最优 133.6% |
| 首 token 延迟 P95秒 · 越低越好 | 1.205最优 | —无数据 | 5.697距最优 372.9% | 2.942距最优 144.2% |
| 首答案 token 延迟中位数秒 · 越低越好 | 45.9距最优 2422.8% | —无数据 | 2.293距最优 26.0% | 1.819最优 |
| 端到端响应中位数秒 · 越低越好 | 57.159距最优 835.4% | —无数据 | 6.111最优 | 10.53距最优 72.3% |
| 上下文窗口tokens · 越高越好 | 128,000距最优 50.0% | 128,000距最优 50.0% | 256,000最优 | 256,000最优 |
| Standard 输入价格百万 tokens · 越低越好 | ¥0.1341$0.0200不可计算 | ——无数据 | ¥0.0000$0.0000最优 | ¥16.7678$2.5000不可计算 |
| Standard 输出价格百万 tokens · 越低越好 | ¥0.6707$0.1000不可计算 | ——无数据 | ¥0.0000$0.0000最优 | ¥67.0712$10.0000不可计算 |
| Standard 缓存输入价格百万 tokens · 越低越好 | ——无数据 | ——无数据 | ——无数据 | ——无数据 |
| terminal bench21 · 越高越好 | 0.019距最优 93.7% | —无数据 | 0.296最优 | —无数据 |
| terminal bench40 · 越高越好 | —无数据 | —无数据 | 0最优 | —无数据 |
人民币价格按服务端汇率 1 USD = 6.7071 CNY 折算,参考日期 2026/9/16。