模型榜单

排名模型厂商能力指数单任务成本输出速度上下文
01Claude Opus 5 (max)Anthropic63$2.3451 t/s1M
02Claude Opus 5 (xhigh)Anthropic63$1.8051 t/s1M
03Claude Fable 5 (with fallback)Anthropic62$3.1466 t/s1M
04Claude Opus 5 (high)Anthropic61$1.2349 t/s1M
05GPT-5.6 Sol (max)OpenAI61$1.2368 t/s1M
06Grok 4.6 (high)xAI61$0.8456 t/s500k
07Kimi K3 (max)Kimi60$0.8438 t/s1.05M
08GPT-5.6 Sol (xhigh)OpenAI59$0.8165 t/s1M
09Claude Opus 5 (medium)Anthropic59$0.7250 t/s1M
10Qwen3.8 MaxAlibaba58$1.1345 t/s1M
11Qwen3.8 2.4T A95BAlibaba58$1.0947 t/s984k
12GPT-5.6 Sol (high)OpenAI57$0.5563 t/s1M
13Muse Spark 1.2 (xhigh)Meta57$0.40暂无1.05M
14GPT-5.6 Terra (max)OpenAI57$0.51120 t/s1M
15Gemini 3.7 Flash (high)Google56$0.40344 t/s1M
16Grok 4.5 (high)xAI56$0.3647 t/s500k
17GPT-5.6 Sol (medium)OpenAI56$0.3760 t/s1M
18Claude Sonnet 5 (max)Anthropic55$1.7272 t/s1M
19Gemini 3.7 Flash (medium)Google53$0.26308 t/s1M
20DeepSeek V4 Pro 0813 (max)DeepSeek53$0.2578 t/s1M
21GPT-5.6 Terra (xhigh)OpenAI53$0.31103 t/s1M
22GLM-5.2 (max)Z.ai53$0.33153 t/s1M
23Claude Opus 5 (low)Anthropic52$0.4348 t/s1M
24GPT-5.6 Luna (max)OpenAI52$0.05168 t/s1M
25Qwen3.8 27BAlibaba52暂无暂无256k
26DeepSeek V4 Flash 0731 (max)DeepSeek52$0.1199 t/s1M
27Gemini 3.6 FlashGoogle52$0.56207 t/s1M
28Gemini 3.7 Flash (low)Google51$0.16282 t/s1M
29GPT-5.6 Sol (low)OpenAI51$0.2357 t/s1M
30GPT-5.6 Terra (high)OpenAI50$0.22106 t/s1M
31GPT-5.6 Luna (xhigh)OpenAI50$0.03172 t/s1M
32Kimi K3 (low)Kimi48$0.2437 t/s1.05M
33Gemini 3.1 Pro PreviewGoogle48$0.33107 t/s1M
34Motif 3Motif Technologies47暂无暂无262k
35GPT-5.6 Luna (high)OpenAI47$0.02148 t/s1M
36GPT-5.6 Terra (medium)OpenAI47$0.12102 t/s1M
37Gemini 3.5 Flash (medium)Google47暂无184 t/s1M
38GPT-5.3 Codex (xhigh)OpenAI46暂无137 t/s400k
39MiniMax-M3MiniMax45$0.1494 t/s1M
40Motif 3 (Beta)Motif Technologies45暂无暂无262k
41DeepSeek V4 Pro (max)DeepSeek45$0.0576 t/s1M
42DeepSeek V4 Pro (high)DeepSeek44$0.0476 t/s1M
43Kimi K2.7 CodeKimi43$0.2238 t/s256k
44MiMo-V2.5-ProXiaomi43$0.0356 t/s1M
45Claude Sonnet 5 (Non-reasoning)Anthropic43$0.4258 t/s1M
46InklingThinking Machines42$0.3483 t/s1M
47Hy3Tencent42$0.0466 t/s256k
48GPT-5.6 Sol (Non-reasoning)OpenAI42$0.2458 t/s1M
49NANex-N2-ProNex AGI42暂无140 t/s262k
50USolar Pro 4Upstage42$0.2251 t/s512k
51GPT-5.6 Terra (low)OpenAI41$0.09105 t/s1M
52Inkling SmallThinking Machines41$0.0763 t/s1M
53JT-4.1 Flash 236B A21BChina Mobile40暂无暂无256k
54SAAgnes 2.5 Pro AlphaSapiens AI40暂无132 t/s1M
55Qwen3.7 PlusAlibaba39$0.2457 t/s1M
56GPT-5.6 Luna (medium)OpenAI39$0.01162 t/s1M
57Nemotron 3 UltraNVIDIA38$0.38133 t/s262k
58MiMo-V2.5Xiaomi38$0.0161 t/s1M
59Ling 3.0 FlashInclusionAI38$0.04暂无262k
60Qwen3.6 27BAlibaba38$0.2956 t/s262k
61Gemini 3.5 Flash-LiteGoogle37$0.10338 t/s1M
62USolar Open2 250BUpstage37暂无暂无1.05M
63MiMo-V2-Omni-0327Xiaomi37暂无暂无256k
64Grok 4.3 (medium)xAI37暂无119 t/s1M
65Grok 4.3 (low)xAI36暂无110 t/s1M
66MiMo-V2-OmniXiaomi36暂无暂无256k
67Gemini 3.5 Flash (minimal)Google36暂无161 t/s1M
68Claude Sonnet 4.6 (Non-reasoning, Low Effort)Anthropic35暂无42 t/s1M
69Muse Glimmer (high)Meta35$0.07111 t/s131k
70STA.X-K2SK Telecom35暂无暂无262k
71GLM-5.2Z.ai35暂无149 t/s1M
72GPT-5.6 Terra (Non-reasoning)OpenAI35$0.10103 t/s1M
73Qwen3.5 397B A17BAlibaba34$0.3699 t/s262k
74MiMo-V2-Flash (Feb 2026)Xiaomi34暂无暂无256k
75LongCat 2.0LongCat34$0.1241 t/s1M
76GPT-5.6 Luna (low)OpenAI34$0.01155 t/s1M
77KAT-Coder-Pro V2KwaiKAT34暂无109 t/s256k
78Qwen3.5 122B A10BAlibaba33$0.25132 t/s262k
79Qwen3.5 397B A17BAlibaba33暂无95 t/s262k
80Qwen3.6 35B A3BAlibaba32$0.29120 t/s262k
81DeepSeek V4 ProDeepSeek32暂无79 t/s1M
82Ring-2.6-1TInclusionAI32$0.36暂无262k
83AG9v3-39A5BAI9Stars32暂无暂无131k
84Qwen3.5 Omni PlusAlibaba31暂无53 t/s256k
85Qwen3.6 27BAlibaba31$0.4056 t/s262k
86o3OpenAI31暂无123 t/s200k
87K-EXAONE 2.0LG AI Research31暂无暂无262k
88Step 3.7 FlashStepFun31$0.09132 t/s262k
89Mistral Medium 3.5Mistral30$0.46144 t/s256k
90Claude 4.5 HaikuAnthropic30$0.2287 t/s200k
91Gemma 4 31BGoogle30$0.0036 t/s256k
92DeepSeek V4 FlashDeepSeek29暂无暂无1M
93GPT-5.5 Instant (June 2026)OpenAI29$0.54119 t/s400k
94JT-35B-FlashChina Mobile29暂无暂无256k
95MiMo-V2.5-ProXiaomi28暂无54 t/s1M
96Qwen3.5 122B A10BAlibaba28$0.20147 t/s262k
97GPT-5.6 Luna (Non-reasoning)OpenAI27$0.01150 t/s1M
98Doubao Seed CodeByteDance Seed26暂无暂无256k
99Gemma 4 26B A4BGoogle26$0.04暂无256k
100Nemotron 3 SuperNVIDIA26$0.23129 t/s1M
101MiMo-V2-FlashXiaomi25暂无暂无256k
102Grok 4.3 (Non-reasoning)xAI25$0.2998 t/s1M
103Qwen3.6 35B A3BAlibaba25$0.63140 t/s262k
104Ling 3.0 TinyInclusionAI25$0.00154 t/s262k
105Qwen3.5 35B A3BAlibaba24$0.24169 t/s262k
106Claude 4.5 HaikuAnthropic24暂无86 t/s200k
107gpt-oss-120b (high)OpenAI24$0.07171 t/s131k
108Nemotron 3.5 LightningNVIDIA24$0.07282 t/s1M
109Command A+Cohere23$0.00206 t/s192k
110K-EXAONELG AI Research22暂无暂无256k
111ERNIE 5.0 Thinking PreviewBaidu22暂无暂无128k
112Gemma 4 31BGoogle22$0.0494 t/s256k
113Gemma 4 12BGoogle22暂无109 t/s256k
114Nova 2.0 Pro Preview (medium)Amazon22暂无114 t/s256k
115Mercury 2Inception22$0.08771 t/s128k
116Qwen3.5 9BAlibaba22$0.2483 t/s262k
117Qwen3 Coder NextAlibaba21$0.34122 t/s256k
118Nova 2.0 Omni (medium)Amazon21暂无暂无1M
119Apriel-v1.6-15B-ThinkerServiceNow21暂无暂无128k
120Nova 2.0 Lite (high)Amazon21暂无148 t/s1M
121Qwen3.5 9BAlibaba21暂无92 t/s262k
122EXAONE 4.5 33BLG AI Research21暂无暂无262k
123Gemma 4 26B A4BGoogle20暂无71 t/s256k
124Qwen3.5 4BAlibaba20暂无51 t/s262k
125North Mini CodeCohere20$0.0026 t/s256k
126Nova 2.0 Pro Preview (low)Amazon20暂无113 t/s256k
127Mistral Small 4Mistral20$0.10158 t/s256k
128Devstral 2Mistral19$0.0035 t/s256k
129Nova 2.0 Lite (medium)Amazon19暂无158 t/s1M
130Qwen3.5 Omni FlashAlibaba19暂无233 t/s256k
131JT-MINIChina Mobile19暂无暂无128k
132Trinity Large ThinkingArcee AI19$0.16180 t/s512k
133HyperNova 60B 2605Multiverse Computing18暂无暂无131k
134Magistral Medium 1.2Mistral18$0.89135 t/s128k
135Nova 2.0 Lite (low)Amazon18暂无159 t/s1M
136Nemotron Cascade 2 30B A3BNVIDIA18暂无暂无1M
137Devstral Small 2Mistral18$0.00132 t/s256k
138MIK2 Think V2MBZUAI Institute of Foundation Models17暂无暂无262k
139LongCat Flash LiteLongCat17暂无暂无256k
140HyperCLOVA X SEED Think (32B)Naver17暂无暂无128k
141K-EXAONELG AI Research17暂无暂无256k
142Qwen3 Next 80B A3BAlibaba17$0.19196 t/s262k
143Nova 2.0 Omni (low)Amazon17暂无暂无1M
144Mi:dm K 2.5 ProKorea Telecom17暂无暂无128k
145AG9v3-3BAI9Stars16暂无暂无131k
146Qwen3.5 4BAlibaba16暂无52 t/s262k
147Mistral Large 3Mistral16$0.0850 t/s256k
148INTELLECT-3Prime Intellect16暂无暂无131k
149USolar Open 100BUpstage15暂无暂无128k
150gpt-oss-20b (high)OpenAI15$0.02129 t/s131k
151Nemotron 3 Nano Omni 30B A3BNVIDIA15暂无335 t/s256k
152gpt-oss-120b (low)OpenAI15$0.02182 t/s131k
153Nemotron 3 NanoNVIDIA15$0.02217 t/s1M
154USolar Pro 3Upstage14$0.15152 t/s128k
155Llama 4 MaverickMeta14$0.04121 t/s1M
156Nova 2.0 Pro PreviewAmazon14暂无106 t/s256k
157gpt-oss-20b (low)OpenAI14暂无176 t/s131k
158MIK2-V2 (high)MBZUAI Institute of Foundation Models14暂无暂无512k
159Qwen3 Next 80B A3BAlibaba14暂无183 t/s262k
160DiffusionGemma 26B A4BGoogle13暂无暂无256k
161Gemma 4 12B (Non-reasoning)Google13暂无112 t/s262k
162Motif-2-12.7BMotif Technologies13暂无暂无128k
163Nova PremierAmazon13暂无34 t/s1M
164MIK2-V2 (medium)MBZUAI Institute of Foundation Models12暂无暂无512k
165Llama Nemotron Super 49B v1.5NVIDIA12暂无113 t/s128k
166Celeris-1Celeris12$0.261466 t/s131k
167Mistral Small 4Mistral12暂无137 t/s256k
168Tri-21B-ThinkTrillion Labs12暂无暂无32k
169Gemma 4 E4BGoogle12暂无51 t/s128k
170MiniCPM5-1BOpenBMB12暂无暂无128k
171SSarvam 105B (high)Sarvam12暂无暂无128k
172Nova 2.0 LiteAmazon12暂无144 t/s1M
173MiniCPM5-1BOpenBMB12暂无暂无128k
174Magistral Small 1.2Mistral11$0.30142 t/s128k
175Ministral 3 14BMistral11$0.1695 t/s256k
176NNanbeige4.1-3BNanbeige11暂无暂无256k
177EXAONE 4.0 32BLG AI Research11暂无暂无131k
178Nova 2.0 OmniAmazon10暂无暂无1M
179Llama 4 ScoutMeta10$0.01134 t/s10M
180Hermes 4 70BNous Research10暂无77 t/s128k
181Falcon-H1R-7BTII UAE10暂无暂无256k
182Gemma 4 E2BGoogle10暂无暂无128k
183Qwen3 Omni 30B A3BAlibaba10暂无99 t/s65.5k
184Step3 VL 10BStepFun9暂无暂无65.5k
185Llama 3.3 70BMeta9暂无88 t/s128k
186Ministral 3 8BMistral9$0.18132 t/s256k
187Llama Nemotron UltraNVIDIA9暂无暂无128k
188ERNIE 4.5 300B A47BBaidu9暂无暂无131k
189Hermes 4 405BNous Research9暂无27 t/s128k
190NVIDIA Nemotron Nano 12B v2 VLNVIDIA9暂无135 t/s128k
191Gemma 4 E4BGoogle9暂无55 t/s128k
192Granite 4.1 30BIBM9暂无暂无131k
193NVIDIA Nemotron Nano 9B V2NVIDIA9暂无180 t/s131k
194Hermes 4 405BNous Research9暂无24 t/s128k
195Nemotron 3 Nano 4BNVIDIA9暂无暂无262k
196Llama Nemotron Super 49B v1.5NVIDIA9暂无113 t/s128k
197MIK2-V2 (low)MBZUAI Institute of Foundation Models8暂无暂无512k
198Kimi Linear 48B A3B InstructKimi8暂无暂无1M
199Llama 3.1 405BMeta8暂无暂无128k
200LFM2.5-8B-A1BLiquid AI8暂无334 t/s32.8k
201Ring-flash-2.0InclusionAI8暂无暂无128k
202Olmo 3.1 32B ThinkAllen Institute for AI8暂无暂无65.5k
203Command ACohere7暂无57 t/s256k
204Qwen3.5 2BAlibaba7暂无暂无262k
205Llama 3.1 Nemotron 70BNVIDIA7暂无130 t/s128k
206Nemotron 3 NanoNVIDIA7暂无223 t/s1M
207NVIDIA Nemotron Nano 9B V2NVIDIA7暂无169 t/s131k
208Ministral 3 3BMistral7$0.13204 t/s256k
209Hermes 4 70BNous Research7暂无66 t/s128k
210Granite 4.1 8BIBM6暂无113 t/s131k
211SSarvam 30B (high)Sarvam6暂无暂无65.5k
212Olmo 3.1 32B InstructAllen Institute for AI6暂无暂无65.5k
213Gemma 4 E2BGoogle6暂无暂无128k
214R1 1776Perplexity6暂无暂无128k
215Llama 3.2 90B (Vision)Meta6暂无暂无128k
216Phi-4 MiniMicrosoft6暂无44 t/s128k
217EXAONE 4.0 32BLG AI Research6暂无暂无131k
218Qwen3.5 2BAlibaba5暂无暂无262k
219Qwen3.5 0.8BAlibaba5暂无暂无262k
220DeepHermes 3 - Mistral 24BNous Research5暂无暂无32k
221Jamba 1.7 LargeAI21 Labs5暂无暂无256k
222Granite 4.0 H SmallIBM5暂无22 t/s128k
223Qwen3 Omni 30B A3BAlibaba5暂无93 t/s65.5k
224LFM2 24B A2BLiquid AI5暂无暂无32.8k
225Phi-4Microsoft5暂无25 t/s16k
226Nova MicroAmazon4暂无276 t/s130k
227Granite 4.1 3BIBM4暂无暂无131k
228NVIDIA Nemotron Nano 12B v2 VLNVIDIA4暂无200 t/s128k
229Phi-4 MultimodalMicrosoft4暂无18 t/s128k
230MiniCPM-V 4.6 1.3BOpenBMB4暂无暂无262k
231Jamba Reasoning 3BAI21 Labs4暂无暂无262k
232Reka Flash 3Reka AI4暂无暂无128k
233Olmo 3 7B ThinkAllen Institute for AI4暂无暂无65.5k
234Molmo 7B-DAllen Institute for AI3暂无暂无4.1k
235Llama 3.2 11B (Vision)Meta3暂无48 t/s128k
236Qwen3.5 0.8BAlibaba3暂无暂无262k
237Exaone 4.0 1.2BLG AI Research3暂无暂无64k
238Olmo 3 7BAllen Institute for AI2暂无暂无65.5k
239Exaone 4.0 1.2BLG AI Research2暂无暂无64k
240LFM2.5-1.2B-ThinkingLiquid AI2暂无暂无32k
241Jamba 1.7 MiniAI21 Labs2暂无暂无258k
242LFM2 2.6BLiquid AI2暂无暂无32.8k
243LFM2.5-1.2B-InstructLiquid AI2暂无暂无32k
244Granite 4.0 H 1BIBM2暂无暂无128k
245Gemma 3 270MGoogle2暂无暂无32k
246Apertus 70B InstructSwiss AI Initiative2暂无暂无65.5k
247Granite 4.0 MicroIBM2暂无暂无128k
248DeepHermes 3 - Llama-3.1 8BNous Research2暂无暂无128k
249Granite 4.0 1BIBM2暂无暂无128k
250Molmo2-8BAllen Institute for AI2暂无暂无36.9k
251LFM2 8B A1BLiquid AI1暂无暂无32.8k
252LFM2.5-VL-1.6BLiquid AI1暂无314 t/s32k
253Granite 4.0 350MIBM1暂无暂无32.8k
254Tiny Aya GlobalCohere1暂无暂无8.19k
255Apertus 8B InstructSwiss AI Initiative1暂无暂无65.5k
256Granite 4.0 H 350MIBM1暂无暂无32.8k

能力指数越高越好;单任务成本越低越好。不同推理档位作为独立配置列出,数据会随榜单实时测量变化。