知识边界

AA-Omniscience

同时观察回答正确性与模型是否知道自己不知道。

数据快照:2026-07-21查看原始来源 ↗
主要指标Omniscience Index
快照日期2026-07-21
447 / 447 个模型
公开评测完整榜单
排名机构模型知识可靠指数准确率幻觉率作答率正确错误未作答
1AnthropicClaude Fable 540.1561.4%54.9%88.6%36811272684
2GoogleGemini 3.1 Pro Preview32.9355.3%49.9%79.4%331513391238
3AnthropicClaude Opus 4.8 (max)27.4346.6%35.9%69.6%279611501824
4SpaceXAIGrok 4.5 (high)26.3852%53.5%78.8%312315401269
5AnthropicClaude Opus 4.7 (max)26.1745.8%36.2%69.5%274711771831
6GoogleGemini 3.5 Flash22.6851.9%60.7%83.2%311317521010
7OpenAIGPT-5.6 Sol (max)21.758.5%88.8%98.3%35122210104
8GoogleGemini 3.5 Flash (medium)21.6551.1%60.3%82.9%306717681028
9OpenAIGPT-5.6 Sol (xhigh)20.5558%89%98.1%34782245114
10OpenAIGPT-5.5 (xhigh)20.0756.9%85.5%96.9%34152211184
11OpenAIGPT-5.6 Sol (high)19.7557.3%88%97.8%34392254135
12OpenAIGPT-5.6 Sol (medium)18.9556.7%87.1%97.5%34012264151
13KimiKimi K318.4246%50.9%76.8%275716521390
14OpenAIGPT-5.6 Sol (low)18.456.4%87.3%97%33862282177
15OpenAIGPT-5.5 (high)18.3756%85.7%97%33632261183
16SpaceXAIGrok 4.3 (high)18.3234.6%25%52.3%20789792863
17MetaMuse Spark 1.1 (xhigh)1840.6%38.1%82.3%243713571062
18OpenAIGPT-5.5 (medium)17.2855.6%86.3%97%33362299183
19SpaceXAIGrok 4.3 (medium)16.9828.4%16%41.4%17076883513
20GoogleGemini 3 Pro Preview (high)15.855.9%90.9%98.6%3354240687
21SpaceXAIGrok 4.20 0309 v215.3527.5%16.7%40.6%16497283565
22AnthropicClaude Sonnet 5 (max)15.3238.3%37.3%64.9%229813792108
23OpenAIGPT-5.5 (low)14.6254.1%86%96.3%32452368220
24AnthropicClaude Opus 4.7 (Non-reasoning, high)14.2243.5%51.9%75.6%261217591461
25AlibabaQwen3.7 Max14.0830.1%22.9%48%18079623120
26SpaceXAIGrok 4.3 (low)13.6725.9%16.5%40%15537333599
27AnthropicClaude Opus 4.6 (max)13.546.4%61.3%81.5%278319731107
28SpaceXAIGrok 4.20 030913.3728.9%21.9%45.1%17359333291
29AnthropicClaude Opus 4.513.2745.7%59.8%81.3%274319471120
30AnthropicClaude Sonnet 4.6 (max)12.3740%46.1%70.8%240216601751
31GoogleGemini 3 Flash11.5754%92.2%98.6%3240254682
32OpenAIGPT-5.5 Instant (May 2026)10.3345.5%64.4%83.1%272821081012
33AlibabaQwen3.6 Max Preview10.237.7%44.2%67.3%226316511963
34OpenAIGPT-5.3 Codex (xhigh)9.8851.8%86.9%96.4%31072514218
35SpaceXAIGrok Build 0.1 06167.4351.4%90.5%97.8%30842638133
36KimiKimi K2.66.4232.8%39.3%60.8%196815832352
37OpenAIGPT-5.4 (xhigh)5.6550%88.6%97%29992660179
38OpenAIGPT-5.1 (high)5.5537.6%51.3%71%225519221739
39XiaomiMiMo-V2-Pro4.9226.8%29.9%49.4%160913143039
40OpenAIGPT-5.4 (low)4.7247.4%81.1%92.4%28432560455
41MetaMuse Spark4.0844.6%73.2%95.3%26772432284
42Z AIGLM-5.2 (max)3.9725.1%28.1%47.2%150312653167
43SpaceXAIGrok 43.7541.4%64.2%79.8%248322581211
44XiaomiMiMo-V2.5-Pro3.622.6%24.5%42.3%135611403465
45OpenAIGPT-5.5 Instant (June 2026)3.4844.1%72.7%86.7%26472438798
46AnthropicClaude Opus 4.6 (high)3.4745.1%76%90.4%27092501574
47AlibabaQwen3.6 Plus2.6526.2%32%51.2%157414152927
48AlibabaQwen3.7 Plus2.3722.2%25.5%43.6%133211903386
49GoogleGemini 3 Pro Preview (low)2.1547.1%84.9%98.2%28242695111
50Thinking MachinesInkling2.0540%63.1%81%239722741141
51Z AIGLM-5226.9%34%53.1%161214922811
52Z AIGLM-5.11.9324.2%29.4%47.5%145213363152
53MiniMaxMiniMax-M31.3715%16.1%30.9%9028204146
54GoogleGemini 3.5 Flash (minimal)1.343.1%73.4%87%25852507782
55OpenAIGPT-5.6 Sol (Non-reasoning)0.8548.2%91.3%97.9%28912840129
56MiniMaxMiniMax-M2.70.6826.1%34.4%53.2%156715262808
57AnthropicClaude 4.5 Sonnet0.3332.4%47.4%67.2%194319231967
58AnthropicClaude 3.7 Sonnet0.3228%38.5%58.1%168216632516
59AnthropicClaude 4 Sonnet0.1822.3%28.5%46.3%134013293223
60OpenAIGPT-5.2 (medium)0.0237.7%60.6%77.4%226422631358
61OpenAIGPT-5.6 Terra (max)-0.2245.9%85.2%94.6%27532766322
62OpenBMBMiniCPM5-1B-0.780.1%0.9%0.9%4515944
63NVIDIANemotron 3 Ultra-0.8321.6%28.5%47.2%129313433166
64OpenAIGPT-5.2 (xhigh)-143.8%79.7%90.7%26282688560
65AnthropicClaude Sonnet 5 (Non-reasoning)-1.1832.6%50.1%69%195520261861
66AnthropicClaude Sonnet 4.6 (Non-reasoning, Low Effort)-1.936.1%59.6%76.3%216822821424
67OpenAIGPT-5.2 Codex (xhigh)-2.4840.7%72.8%85.8%24422591850
68AnthropicClaude Sonnet 4.6 (Non-reasoning)-2.9338%65.9%81.5%227824541108
69OpenAIGPT-5.6 Terra (xhigh)-3.344.7%86.7%95.3%26812879281
70GoogleGemini 3 Flash-3.6245.5%90.2%98.2%27312948107
71AnthropicClaude Opus 4.5-3.9540.7%75.4%88.3%24442681702
72CohereCommand A+-3.988.8%14.1%22.3%5317704665
73OpenAIGPT-5.6 Terra (high)-4.0744.5%87.5%95.5%26692913269
74AnthropicClaude 4.5 Haiku-4.2217.4%26.2%41.5%104612993512
75OpenAIGPT-5.5 (Non-reasoning)-5.0545.1%91.2%97.7%27043007137
76OpenAIGPT-5.6 Terra (medium)-5.1744%87.9%95.7%26412951256
77OpenAIGPT-5.1 Codex (high)-6.0239.2%74.4%87.7%23522713739
78SpaceXAIGrok 3 mini Reasoning (high)-6.0515.4%25.4%37.5%92412873751
79Z AIGLM-5.2-6.420.3%33.5%48%121916033120
80OpenAIGPT-5 Codex (high)-6.7738.7%74.1%86.1%23202726837
81OpenAIGPT-5.6 Terra (low)-7.1243%87.8%95.6%25773004266
82AnthropicClaude 4.5 Haiku-7.6713.7%24.7%37%82012803777
83OpenAIGPT-5 (high)-8.0840.6%82.1%90.6%24392924562
84KimiKimi K2.5-8.1234.3%64.6%79.8%205925461211
85AnthropicClaude 4.5 Sonnet-9.227.5%50.7%68.7%165122031876
86AnthropicClaude 3.7 Sonnet-9.2328.3%52.3%67.9%169822521924
87AnthropicClaude 4 Sonnet-9.2322.4%40.8%56.4%134619002617
88XiaomiMiMo-V2.5-9.3317.3%32.2%44.4%103615963336
89DeepSeekDeepSeek V4 Pro (high)-9.741.8%88.6%94.8%25103092313
90KimiKimi K2.6-9.9723.1%42.9%59.4%138419822435
91DeepSeekDeepSeek V4 Pro (max)-10.0243.3%94%98.2%25973198109
92OpenAIGPT-5 (low)-10.0237.4%75.8%88.2%22462847706
93OpenAIGPT-5 (medium)-10.0838.9%80.1%91.1%23322937536
94China MobileJT-4.1 Flash 236B A21B-10.1223.1%43.3%59.3%138819952442
95OpenAIo1-10.5534.7%69.3%81.2%208127141130
96KimiKimi K2.7 Code-10.738.6%80.3%90.6%23162958561
97OpenAIGPT-4o (Nov)-10.7519.7%37.9%51.7%118118262897
98OpenAIGPT-5 mini (medium)-10.7822.2%42.4%58.8%133219792474
99OpenAIGPT-5.6 Luna (max)-11.2341.5%90.1%96.9%24903164184
100Z AIGLM-5-11.2322.7%43.9%59.5%136120352429
101OpenAIGPT-5.6 Luna (xhigh)-11.641.3%90.1%96.8%24783174191
102AlibabaQwen3.5 Omni Plus-12.317.4%36%48.9%104517833069
103OpenAIGPT-5.6 Luna (high)-12.4540.7%89.5%96.6%24403187206
104OpenAIGPT-5.2-13.7229.3%60.8%75.4%175725801477
105OpenAIGPT-5.6 Luna (medium)-13.8239.6%88.5%95.6%23783207263
106KimiKimi K2.5-1422.9%47.9%64.3%137422142142
107XiaomiMiMo-V2-Omni-0327-14.0218.5%39.9%51.7%110919502900
108GoogleGemini 2.5 Pro-14.339%87.4%95.8%23403198251
109Z AIGLM-5-Turbo-15.0829%62.2%74.6%174226471525
110OpenAIGPT-5.6 Luna (low)-15.238.6%87.7%95.1%23183230295
111OpenAIo3-15.2738.4%87.1%93.9%23043220365
112GoogleGemini 3.1 Flash-Lite-15.5236.4%81.6%91.1%21843115533
113OpenAIGPT-5.4 (Non-reasoning)-15.8836.8%83.2%91.9%22053158484
114DeepSeekDeepSeek V3.2 Speciale-16.4838.4%88.9%94.6%23013290325
115OpenAIGPT-5.1 Codex mini (high)-16.5222.7%50.8%63.9%136323542166
116OpenBMBMiniCPM5-1B-16.621.7%18.7%20.2%10411014790
117OpenAIGPT-5 mini (high)-17.1823.9%54.1%68.9%143724681866
118Motif TechnologiesMotif 3 (Beta)-17.2821%48.5%60.6%126222992367
119MetaLlama 3.1 405B-17.322.3%51%64.6%134023782126
120XiaomiMiMo-V2-Omni-17.4318.7%44.4%61.7%112121672299
121XiaomiMiMo-V2-Flash (Feb 2026)-18.4720.2%48.4%59.6%121023182424
122TencentHy3-18.531.5%73%83.7%18903000978
123OpenAIGPT-5.4 mini (xhigh)-18.6837.5%89.8%95.3%22483369283
124OpenAIGPT-5.4 nano-18.8221%50.4%62.5%126023892247
125Z AIGLM 5V Turbo-18.9829.1%67.9%79.7%174828871218
126AlibabaQwen3.6 27B-19.7819.2%48.3%59.4%115323402435
127GoogleGemma 4 E4B-20.058.6%31.3%37.9%51317163726
128OpenAIGPT-5.4 mini (medium)-20.1236.5%89.2%94.8%21903397313
129KimiKimi K2 Thinking-20.4830%72.2%84.6%18013030922
130DeepSeekDeepSeek V3.2-20.8833.5%81.7%89.8%20083261613
131Z AIGLM-5.1-21.425.7%63.4%73.7%154328271581
132AlibabaQwen3.6 35B A3B-21.4318.9%49.7%60.3%113324192381
133KwaiKATKAT-Coder-Pro V2-21.622%55.8%66.8%131826141995
134OpenAIGPT-4o (Aug)-21.7223.7%59.5%70.3%142027231785
135AlibabaQwen3 Coder 480B-22.214.7%43.3%55.9%88322152649
136DeepSeekDeepSeek V4 Flash (high)-22.2835.5%89.7%94.5%21323469327
137LongCatLongCat 2.0-22.6329.8%74.7%83.8%17883146974
138DeepSeekDeepSeek V4 Flash (max)-22.937.2%95.8%99.1%2233360757
139AnthropicClaude 3.5 Haiku-23.1813%41.5%50.4%77821692976
140China MobileJT-35B-Flash-23.2524.5%63.3%74%147128661559
141OpenAIGPT-5.6 Terra (Non-reasoning)-23.4736.2%93.5%98.4%2172358097
142GoogleGemma 4 E2B-23.986.7%32.9%38.4%40118403694
143DeepSeekDeepSeek V3.1 Terminus-24.3227.9%72.5%83%167631351023
144OpenAIGPT-5 nano (medium)-24.8816.1%48.8%63%96624592217
145Deep CogitoCogito v2.1-24.9830.2%79.1%87.6%18123311744
146OpenAIGPT-5.6 Luna (Non-reasoning)-25.1227.9%73.5%83.5%16733180990
147Z AIGLM-4.5-25.4524.9%67.2%79.4%149730241235
148MistralMagistral Medium 1.2-26.2320.8%59.5%69.9%125128251808
149MistralMagistral Medium 1-26.4220.7%59.4%68.8%124328281871
150KimiKimi K2 0905-26.4725.4%69.6%79%152631141259
151Z AIGLM-4.6V-26.4815.6%49.9%60.8%93725262351
152DeepSeekDeepSeek R1 0528-27.0830.9%84%90.3%18573482584
153KimiKimi K2-27.5226.8%74.2%84.6%16093260922
154Nex AGINex-N2-Pro-27.6234.4%94.5%98.4%2064372198
155OpenAIGPT-5 nano (high)-27.7518.3%56.3%68.1%109727621912
156GoogleGemini 2.5 Flash-27.8525.1%70.7%83.7%15053176980
157DeepSeekDeepSeek V3.1-28.3728.8%80.3%89.1%17293431653
158SpaceXAIGrok 4 Fast-28.422.6%66%76.4%135830621413
159SpaceXAIGrok 4.1 Fast-28.7325.3%72.4%80.8%152032441155
160OpenAIGPT-5.4 nano (xhigh)-29.5525.4%73.6%81.8%152232951093
161AlibabaQwen3.5 397B A17B-29.7831.4%89.1%93.8%18813668374
162MistralMistral Small 4-29.922.1%66.8%75.3%132731211484
163DeepSeekDeepSeek V3.2 Exp-29.9328.2%81.1%87.3%16953491765
164DeepSeekDeepSeek V4 Pro-30.0830.8%87.9%93.8%18463651371
165MetaLlama 3.1 8B-30.338.1%41.8%49.3%48423043041
166GoogleGemma 3 270M-31.320.9%32.5%35%5219313898
167DeepSeekDeepSeek R1 (Jan)-31.3230.7%89.5%94.6%18433722326
168MistralMistral Medium 3-31.4718.3%60.9%69.7%109829861817
169Z AIGLM-4.6-31.620.8%66.1%76.5%124731431412
170MistralDevstral Medium-31.6219.1%62.6%71%114330401738
171SpaceXAIGrok 4.3 (Non-reasoning)-32.324.2%74.5%81.7%145033881098
172SpaceXAIGrok 3-32.3328.3%84.6%93.5%16993639392
173MiniMaxMiniMax-M2.1-32.8320.5%67%77.2%122831981368
174Nous ResearchHermes 4 405B-3326.3%80.4%87.4%15763556756
175Liquid AILFM2.5-8B-A1B-33.279.3%47%52.6%56125572845
176MBZUAI Institute of Foundation ModelsK2 Think V2-33.9215.7%58.9%66.8%94329781991
177OpenAIGPT-5 (minimal)-33.9528.1%86.2%94.6%16843721325
178MistralMistral Large 2 (Nov)-34.0520.1%67.8%75.6%120832511462
179AlibabaQwen3 Max Thinking-34.4229.9%91.8%97.4%17953860155
180ByteDance SeedDoubao Seed Code-34.4325.1%79.4%86%15043570841
181Z AIGLM-4.7-34.629.3%90.3%96.8%17563832193
182TencentHy3-preview-34.6328%86.9%92.4%16793757455
183GoogleGemini 2.5 Flash (Sep)-34.7228.9%89.5%93.9%17343817368
184OpenAIGPT-5.1-34.9529.1%90.2%95%17443841299
185AlibabaQwen3.5 0.8B-35.181.1%36.7%38.5%6521763693
186OpenAIo4-mini (high)-35.7524.6%80.1%86.2%14783623830
187BaiduERNIE 4.5 300B A47B-35.8818.7%67.1%75.8%112132741449
188SpaceXAIGrok Code Fast 1-3623.8%78.5%85.2%14293589891
189Nous ResearchHermes 4 405B-36.0530.2%94.8%98.1%18103973114
190AlibabaQwen3.5 397B A17B-36.0824.3%79.8%88.6%14603625683
191OpenAIGPT-4.1-36.1824.2%79.6%97.9%14513622128
192AmazonNova Premier-36.2719.1%68.5%80.7%114733231158
193MistralMistral Medium 3.5-36.3225.1%82%87.4%15063685757
194TencentHy3-preview-36.422.4%75.7%83.8%13413525971
195KwaiKATKAT-Coder-Pro V1-37.4317.5%66.6%74.4%104932951538
196StepFunStep 3.7 Flash-37.5525.4%84.4%90.2%15243777588
197DeepSeekDeepSeek V3 0324-37.5722.2%76.7%91.1%13303584532
198InclusionAIRing-2.6-1T-37.7525.1%83.8%89.7%15043769618
199AlibabaQwen3 Max Thinking (Preview)-37.8328%91.3%94.8%16773947312
200Z AIGLM-4.6V-37.8717.5%67.1%74.5%105033221533
201GoogleGemini 2.5 Flash (Sep)-38.1226.7%88.3%94.9%15993886307
202XiaomiMiMo-V2.5-Pro-38.327%89.3%94.3%16173915345
203OpenAIGPT-5.4 nano-38.5813.9%60.9%68.6%83231471885
204MistralMistral Large 3-39.4324.1%83.7%91.8%14473813493
205AlibabaQwen3.5 122B A10B-39.5824.7%85.5%91.4%14843859516
206MiniMaxMiniMax-M2.5-39.726.2%89.3%92.8%15723954429
207AmazonNova Lite-40.429.7%55.6%62.6%58430092242
208AlibabaQwen3 Max (Preview)-40.5824.3%85.7%92.6%14573892445
209MetaLlama 3.1 70B-40.8519.4%74.7%84.4%11633614934
210StepFunStep 3.5 Flash-40.923.9%85.2%90.6%14343888564
211NVIDIALlama 3.1 Nemotron 70B-41.0516.4%68.8%78.5%98734501289
212DeepSeekDeepSeek V3.1-41.0723.1%83.5%90.8%13883852553
213OpenAIo3-mini (high)-41.3321%79%86.6%12623742806
214DeepSeekDeepSeek V3 (Dec)-41.3525.4%89.4%94%15224003357
215GoogleGemma 4 E4B-41.528.3%54.4%58.9%49929902467
216Z AIGLM-4.6-41.7327.3%95%97.5%16394143151
217MetaLlama 4 Maverick-41.7724.3%87.3%93.7%14593965380
218DeepSeekDeepSeek V3.1 Terminus-41.8823.4%85.3%92.2%14073920469
219NanbeigeNanbeige4.1-3B-41.939.6%57%62.5%57530912248
220GoogleGemini 2.5 Flash-42.0226.5%93.3%96.6%15924113204
221AlibabaQwen3.5 27B-42.0221%79.7%86.3%12583779824
222NVIDIANVIDIA Nemotron 3 Super-42.0724%87%91.4%14423966516
223AlibabaQwen3 235B 2507-42.116.8%70.8%83.4%10073533995
224GoogleGemini 2.0 Flash-42.5523.6%86.5%91.3%14153968519
225GoogleGemini 2.5 Flash-Lite (Sep)-42.6714.3%66.4%72%85734171678
226SpaceXAIGrok 4.20 0309-42.6826%92.9%96.5%15614122210
227NVIDIANVIDIA Nemotron Nano 9B V2-42.8711.3%61%66.2%67632482026
228AlibabaQwen3 Max-43.1224.4%89.4%94%14664053363
229MistralMagistral Small 1-43.3718.3%75.5%81.3%109837001120
230XiaomiMiMo-V2-Flash-43.4524.9%91.1%96.4%14954102215
231Z AIGLM-4.5V-43.5720.6%80.9%90%12373851599
232StepFunStep 3.5 Flash 2603-43.6725%91.6%94.8%15024122309
233Allen Institute for AIOlmo 3.1 32B Think-44.1213.6%66.8%72.9%81634631629
234BaiduERNIE 5.0 Thinking Preview-44.1321.9%84.6%91.3%13153963525
235DeepSeekDeepSeek V4 Flash-44.1826.1%95.1%97.9%15664217127
236Arcee AITrinity Large Thinking-44.1822.8%86.6%90.6%13654016563
237AlibabaQwen3 235B-44.5818.4%77.2%83%110437791019
238SpaceXAIGrok 4.20 0309 v2-44.626.6%96.9%99.5%1594427033
239AlibabaQwen3 VL 235B A22B-44.6220.8%82.5%89.1%12463923651
240GoogleGemini 2.5 Flash-Lite-44.6317.5%75.4%82.8%105337311030
241GoogleGemma 4 31B-45.4219.9%81.6%89.7%11953920618
242MistralMistral Medium 3.1-45.4719.8%81.3%90%11863914598
243NVIDIALlama Nemotron Ultra-45.5219.9%81.7%86.8%11943925792
244AlibabaQwen3 235B A22B 2507-45.5523.1%89.2%93.8%13854118371
245NVIDIALlama Nemotron Super 49B v1.5-45.6312.2%65.9%72.6%73234701643
246AmazonNova 2.0 Pro Preview (low)-45.7722.1%87.2%91.5%13284074509
247NVIDIALlama Nemotron Super 49B v1.5-45.8216.9%75.5%82.3%101637651064
248OpenAIGPT-5.4 mini-45.8525.3%95.3%98.1%15204271113
249MistralDevstral 2-46.1721%85.1%90%12614031602
250Z AIGLM-4.7-46.2823.9%92.3%96.1%14354212236
251AlibabaQwen3.5 35B A3B-46.3820.4%84%89.6%12274010626
252DeepSeekDeepSeek R1 Distill Llama 70B-46.4519.3%81.5%86.9%11593946787
253DeepSeekDeepSeek V3.2-46.7324.2%93.5%96.1%14504254232
254AlibabaQwen3.5 27B-46.7515.7%74.1%79.7%94437491218
255MiniMaxMiniMax-M2-46.9322%88.4%94.8%13214137315
256DeepSeekDeepSeek V3.2 Exp-47.0822.7%90.3%95%13624187299
257Nous ResearchHermes 4 70B-47.3718.2%80.1%87.9%10913933724
258MiniMaxMiniMax M1 80k-47.4221.1%86.8%95%12654110302
259AlibabaQwen3.5 2B-47.484.7%54.7%58.6%28131302485
260AmazonNova Pro-47.5817%77.9%83.3%102238771005
261AnthropicClaude 3 Haiku-47.6517.2%78.2%86.1%10303889837
262InclusionAIRing-1T-47.8719.8%84.4%92.3%11884060462
263AmazonNova 2.0 Pro Preview (medium)-48.0522%89.8%93.9%13194202368
264GoogleGemma 4 26B A4B-48.0718.1%80.9%90.5%10893973570
265MBZUAI Institute of Foundation ModelsK2-V2 (low)-48.0716.1%76.4%81.4%96538491116
266AmazonNova 2.0 Pro Preview-48.1516.1%76.5%84.5%9653854929
267MistralMistral Small 4-48.3716.3%77.2%83.3%97538771003
268CohereCommand A-48.4715.7%76.2%83.2%94338511006
269XiaomiMiMo-V2-Flash-48.4815.2%75.1%81.5%91338221111
270AmazonNova Micro-48.679.9%65%69%59335131857
271CohereNorth Mini Code-49.0718.6%83.1%87.5%11144058750
272AlibabaQwen3 14B-49.2314.9%75.5%80.5%89738511168
273NVIDIALlama 3.3 Nemotron Super 49B-49.6813.5%73%78.3%80837891300
274SpaceXAIGrok 4 Fast-49.8715.9%78.2%91.5%9553947509
275Nous ResearchHermes 4 70B-49.9223.3%95.4%97.9%13974392124
276MBZUAI Institute of Foundation ModelsK2-V2 (medium)-49.9717.4%81.5%86.3%10424040824
277AmazonNova 2.0 Omni (low)-49.9818.5%84%88.8%11084107674
278OpenAIgpt-oss-120b (high)-50.0521.5%91.2%94.2%12914294346
279OpenAIGPT-4.1 mini-50.1317.5%82%96.2%10504058228
280OpenAIgpt-oss-120b (low)-50.1515.7%78.1%95%9433952300
281AlibabaQwen3 VL 32B-50.216.8%80.4%87.5%10054017747
282MistralMistral Small 3.1-50.3215%76.8%83%90039191023
283MistralMistral Small 3.2-50.4214.7%76.3%82.2%88039051067
284AlibabaQwen3 32B-50.5317.3%82%86.7%10364068797
285AlibabaQwen3 Next 80B A3B-50.5717.4%82.2%91.6%10424076506
286AlibabaQwen3 VL 235B A22B-50.6720.2%88.7%94.8%12104250310
287AlibabaQwen3 Coder 30B A3B-50.8515.8%79.2%83.8%9504001974
288ByteDance SeedSeed-OSS-36B-Instruct-50.8517.8%83.4%90.2%10654116587
289SpaceXAIGrok 4.1 Fast-50.9517%81.8%86.8%10184075790
290InclusionAILing-2.6-1T-5121.4%92.2%96.4%12874347215
291Prime IntellectINTELLECT-3-51.0219.2%87%91.3%11544215522
292AmazonNova 2.0 Lite (low)-51.0715.9%79.6%91%9524016542
293Allen Institute for AIOlmo 3.1 32B Instruct-51.0811.3%70.4%77.7%68037451336
294MistralDevstral Small-51.1514.7%77.1%82.2%88039491071
295AlibabaQwen3 30B-51.216.1%80.3%84.9%9684040905
296GoogleGemma 4 31B-51.3516.9%82.1%86.9%10124093786
297LG AI ResearchEXAONE 4.5 33B-51.4816.2%80.7%84.5%9704059933
298NVIDIANVIDIA Nemotron 3 Nano-51.6317.1%82.9%87.3%10274125765
299GoogleGemma 4 12B-51.8816%80.8%84.9%9604073907
300MetaLlama 3.3 70B-51.9217.9%85.1%94.2%10764191346
301AlibabaQwen3 VL 8B-52.0520.2%90.4%94.6%12094332324
302Liquid AILFM2 2.6B-52.15.2%60.5%65.9%31434402048
303AlibabaQwen2.5 72B-52.2217.6%84.7%89.5%10554188629
304InceptionMercury 2-52.2820.5%91.5%94.6%12294366324
305MetaLlama 4 Scout-52.3714.5%78.3%84.3%8734015940
306AlibabaQwen3 VL 8B-52.4220.1%90.7%94.8%12054350314
307NVIDIANemotron Cascade 2 30B A3B-52.4717.6%85%89.2%10564204646
308AlibabaQwen3.5 9B-52.5215.9%81.4%85.3%9544105881
309GoogleGemma 4 12B (Non-reasoning)-52.7312%73.5%77.8%71838821333
310AlibabaQwen3.6 27B-52.7817%84%88%10194186721
311NaverHyperCLOVA X SEED Think (32B)-52.8714.6%79.1%85.6%8784050862
312AlibabaQwen3 235B-52.9718.2%87%91.4%10924270518
313GoogleGemini 2.5 Flash-Lite (Sep)-53.0818%86.8%91.1%10824267533
314Z AIGLM-4.5V-53.5217.9%87%93.8%10744285374
315MetaLlama 3 70B-53.5817.6%86.5%91.9%10594274487
316UpstageSolar Pro 3-53.7818.1%87.8%91.7%10874314499
317AlibabaQwen3.5 122B A10B-53.918.6%89.1%94.8%11164350311
318UpstageSolar Open 100B-54.118.1%88.1%92.6%10854331443
319AmazonNova 2.0 Lite (high)-54.1218.8%89.8%93.7%11294376377
320OpenAIGPT-5 mini (minimal)-54.1218.2%88.4%93.2%10914338411
321Swiss AI InitiativeApertus 70B Instruct-55.113.2%78.6%85.2%7904096891
322MetaLlama 3.2 1B-55.226.9%66.8%69.8%41637291810
323Trillion LabsTri-21B-think Preview-55.288.7%70.1%75.4%52438411475
324InclusionAILing-1T-55.4819.9%94%97.5%11914520151
325AI21 LabsJamba 1.7 Large-55.5519.7%93.8%97.8%11834516130
326AmazonNova 2.0 Lite (medium)-55.6218.3%90.4%94.3%10974434342
327ServiceNowApriel-v1.5-15B-Thinker-55.7816.1%85.7%89.4%9664313635
328NVIDIANemotron 3 Nano Omni 30B A3B Reasoning-56.0314.8%83.1%86.5%8884250809
329OpenAIGPT-4.1 nano-56.4213.3%80.4%86.5%7984183810
330Multiverse ComputingHyperNova 60B 2605-56.4314.1%82.2%86.4%8494235819
331UpstageSolar Pro 2-56.5719.5%94.5%97.3%11694563162
332MBZUAI Institute of Foundation ModelsK2-V2 (high)-56.718.3%91.8%94.5%10974499332
333MicrosoftPhi-4-56.7213.2%80.5%86.1%7914194832
334MistralDevstral Small 2-56.7315.3%85.1%91.2%9204324528
335AlibabaQwen3.5 4B-56.7712.8%79.7%84.4%7674173934
336AlibabaQwen3 30B A3B 2507-56.8716.2%87.3%90.4%9744386576
337MistralDevstral Small (May)-56.9815.8%86.6%90.1%9514370593
338GoogleGemini 2.5 Flash-Lite-5715.3%85.4%91.1%9194339536
339NVIDIANVIDIA Nemotron Nano 9B V2-57.19.5%73.7%78.6%57239981286
340Korea TelecomMi:dm K 2.5 Pro-57.2716.8%88.9%95.9%10054441249
341AlibabaQwen3 VL 30B A3B-57.3216.9%89.2%92.8%10114450435
342AmazonNova 2.0 Omni (medium)-57.8317.9%92.2%95.3%10734543280
343LG AI ResearchK-EXAONE-57.9316.5%89.1%91.7%9894465497
344NVIDIALlama 3.3 Nemotron Super 49B-58.3316.9%90.5%93.9%10144514367
345Allen Institute for AIOlmo 3 32B Think-58.414.4%85.2%89.8%8674371610
346ServiceNowApriel-v1.6-15B-Thinker-58.5317%90.9%94.3%10174529345
347InclusionAIRing-flash-2.0-58.6716.3%89.6%92.7%9784498437
348StepFunStep3 VL 10B-5912.7%82.2%85.9%7634303846
349GoogleDiffusionGemma 26B A4B-59.0316.8%91.1%93.8%10084550373
350Liquid AILFM2 24B A2B-59.076.4%70%72.4%38539291655
351AmazonNova 2.0 Lite-59.113.5%83.9%89.6%8104356626
352AlibabaQwen3 Next 80B A3B-59.117.6%93.2%95.4%10584604275
353Z AIGLM-4.7-Flash-59.3315.9%89.5%94.5%9544514328
354SarvamSarvam 105B (high)-59.5317.6%93.5%96.1%10544626234
355AlibabaQwen3 4B 2507-59.6310.3%78%84.3%6194197943
356AlibabaQwen3.6 35B A3B-59.8317.1%92.7%95.3%10244614285
357OpenAIgpt-oss-20b (low)-60.1314.2%86.6%89.8%8514459609
358AlibabaQwen3 Coder Next-60.7215.8%90.9%95.7%9494592259
359AlibabaQwen3 Omni 30B A3B-60.7214.8%88.7%91.5%8894532509
360LG AI ResearchEXAONE 4.0 32B-60.813.6%86.1%89.7%8164464617
361AlibabaQwen3 4B 2507-60.8212.7%84.1%86.8%7594408793
362IBMGranite 4.0 H Small-60.9214.4%87.9%90.7%8624517558
363MicrosoftPhi-4 Mini-61.028.3%75.5%81.8%49641571093
364LG AI ResearchK-EXAONE-61.1712.2%83.5%93.9%7314401364
365Motif TechnologiesMotif-2-12.7B-61.1815.3%90.4%93%9214592422
366GoogleGemma 4 E2B-61.378%75.4%78.1%47941611311
367UpstageSolar Pro 2-61.6815.6%91.5%96.4%9344635218
368AlibabaQwen3.5 35B A3B-6215.7%92.1%96.2%9404660231
369Korea TelecomMi:dm K 2.5 Pro Preview-62.0216.1%93.1%96.3%9654686222
370GoogleGemma 4 26B A4B-62.0315.7%92.2%94.8%9434665310
371TII UAEFalcon-H1R-7B-62.0814.1%88.7%92%8454570481
372Z AIGLM-4.7-Flash-62.2312%84.3%96.9%7194453186
373SarvamSarvam M-62.2514.8%90.5%93.2%8904625407
374LG AI ResearchEXAONE 4.0 32B-62.310.3%81%86%6214359842
375Z AIGLM-4.5-Air-62.5215.5%92.3%95.5%9304681271
376AlibabaQwen3 VL 32B-62.8214.9%91.2%93.9%8924661365
377AlibabaQwen3 VL 30B A3B-6315.5%92.9%95.7%9314711259
378MetaLlama 3.2 11B (Vision)-63.0810.2%81.6%85.6%6124397866
379Trillion LabsTri-21B-Think-63.312%85.5%87.9%7184516724
380InclusionAILing-flash-2.0-63.3714%90%93.2%8404642408
381OpenAIgpt-oss-20b (high)-63.9215.5%94.1%96.4%9324767215
382OpenAIGPT-5 nano (minimal)-63.9811%84.3%91.7%6624501500
383MistralMinistral 3 3B-64.17.6%77.5%82.5%45443001051
384DeepSeekDeepSeek R1 0528 Qwen3 8B-64.1212%86.5%89.4%7204567637
385NVIDIANVIDIA Nemotron Nano 12B v2 VL-64.4213.8%90.7%94.6%8264691322
386AmazonNova 2.0 Omni-64.512.9%88.8%91.8%7724642489
387China MobileJT-MINI-64.5813.9%91.2%93.8%8364711370
388IBMGranite 4.1 8B-64.612.1%87.3%90.2%7264602588
389Reka AIReka Flash 3-64.713.4%90.1%93.2%8024684407
390MistralMistral 7B-64.777.9%79%84.6%4764362922
391MistralMagistral Small 1.2-64.9813.5%90.7%93.7%8094708376
392AlibabaQwen3 8B-65.0712.8%89.3%93.5%7674671392
393AlibabaQwen3 30B-65.0711.7%86.8%92.5%6994603451
394AlibabaQwen3.5 Omni Flash-65.6314.4%93.6%95.9%8674805248
395InclusionAILing 2.6 Flash-65.6815.4%95.8%99.1%923486452
396GoogleGemma 3 27B-65.8812.5%89.5%95%7494702303
397AlibabaQwen3 30B A3B 2507-65.9215.1%95.4%97.3%9064861164
398MistralMinistral 3 14B-66.8212.3%90.2%95.5%7384747273
399AlibabaQwen3 14B-67.3812.7%91.8%95%7644807302
400IBMGranite 4.1 30B-67.7814%95.1%96.9%8394906184
401MistralMinistral 3 8B-67.8311.2%89%96.8%6714741194
402Allen Institute for AIMolmo2-8B-69.1211.3%90.7%93.2%6794826410
403AlibabaQwen3 Omni 30B A3B-69.1814.4%97.7%99.2%865501650
404NVIDIANVIDIA Nemotron 3 Nano-69.2211.4%90.9%93.6%6824835385
405AlibabaQwen3 VL 4B-69.3512.2%92.8%94.3%7304891344
406MetaLlama 3 8B-69.559.8%88%92.6%5884761443
407LongCatLongCat Flash Lite-70.113.1%95.7%98%7864992122
408AlibabaQwen3.5 9B-70.7514.1%98.8%99.7%846509119
409NVIDIANVIDIA Nemotron 3 Nano 4B-71.539.4%89.4%91%5674859539
410Liquid AILFM2 1.2B-71.624.3%79.3%81.8%25745541095
411SarvamSarvam 30B (high)-72.0512.7%97%98.2%7605083106
412IBMGranite 4.0 350M-72.153.2%77.8%82.1%19045191072
413NVIDIANVIDIA Nemotron Nano 12B v2 VL-72.7511.4%94.9%97.4%6825047154
414IBMGranite 4.0 H 1B-73.655.3%83.4%85.1%3204739895
415Allen Institute for AIOlmo 3 7B Think-73.6810.8%94.7%95.7%6495070257
416AI21 LabsJamba Reasoning 3B-73.686.6%86%88.1%3964817713
417Liquid AILFM2.5-1.2B-Instruct-73.786%84.8%89%3594786658
418AlibabaQwen3 8B-74.110.3%94.2%97.7%6215067138
419AI21 LabsJamba 1.7 Mini-74.2511.4%96.7%98%6845139119
420Swiss AI InitiativeApertus 8B Instruct-74.7210.3%94.7%97.2%6155098171
421AlibabaQwen3.5 4B-74.7211.4%97.2%98.2%6855168108
422Liquid AILFM2 8B A1B-75.136.8%88%94.2%4104918347
423AlibabaQwen3 VL 4B-75.8210.3%96.1%98.5%621517088
424IBMGranite 3.3 8B-76.628.5%93.1%98.5%512510991
425GoogleGemma 3 12B-76.9810.2%97.1%99.2%613523248
426AlibabaQwen3 1.7B-77.28.4%93.4%96%5045136238
427GoogleGemma 3 1B-77.273.6%83.9%86.5%2154851811
428IBMGranite 4.1 3B-77.389.3%95.5%96.5%5555198207
429Allen Institute for AIOlmo 3 7B-77.757.3%91.8%92.7%4405105436
430IBMGranite 4.0 Micro-77.929%95.5%97.1%5415216172
431InclusionAILing-mini-2.0-78.778.7%95.8%97.5%5225248151
432GoogleGemma 3n E2B-79.586.7%92.4%95.3%3995174279
433GoogleGemma 3n E4B-80.47.8%95.6%98.7%465528979
434AlibabaQwen3 0.6B-81.785.5%92.4%94.5%3315238330
435IBMGranite 4.0 1B-81.826%93.5%95.7%3635272257
436LG AI ResearchExaone 4.0 1.2B-82.436.2%94.5%96.2%3715317231
437LG AI ResearchExaone 4.0 1.2B-82.584.7%91.5%93.3%2795234404
438AlibabaQwen3.5 2B-82.837.4%97.4%98.6%444541486
439GoogleGemma 3 4B-83.327.8%98.8%99.7%468546720
440AlibabaQwen3 1.7B-83.476.6%96.4%98.8%396540473
441Liquid AILFM2.5-VL-1.6B-83.875.2%94%97.5%3125344149
442Liquid AILFM2.5-1.2B-Thinking-83.876.6%96.9%98%3975429121
443OpenBMBMiniCPM-V 4.6 1.3B-84.676.2%96.8%98.3%3715451100
444CohereTiny Aya Global-84.725.7%95.8%97.6%3395422146
445AlibabaQwen3 0.6B-85.534.2%93.7%95.6%2535385262
446IBMGranite 4.0 H 350M-87.253.7%94.4%94.7%2205455319
447AlibabaQwen3.5 0.8B-894.8%98.6%99.5%290563029

这个评测关注什么

同时观察回答正确性与模型是否知道自己不知道。

阅读时注意

评测结果受任务集合、评分方法和模型版本影响。本站保留原始口径,不用它生成新的综合排名。

适用场景

用于回答一个明确的问题:模型在这一类任务上的相对表现如何,而不是“哪个模型绝对最好”。