最佳推理 AI 模型

279 个模型 · 推理、数学和知识的平均值。至少两项结果。方法

上榜模型

2026年11月

#模型推理指数输出 / 1M速度

01Claude Fable 5.284.385.6$40.0064/s

02Gemini 4 Carbon80.282.7$10.0031/s

03GPT-6 Astra80.178.0$50.0034/s

04GPT-6.1 Sol79.378.0$10.0034/s

04Claude Opus 5.579.378.6$20.0076/s

06Gemini 4 Argon77.478.5$10.0028/s

07Claude Fable 5.176.776.2$50.0069/s

08GPT-6.1 Terra76.277.1$5.0046/s

09Muse Wonder Full75.677.3——

10GPT-5.5 Pro75.574.8$180.0019/s

11Claude Fable 575.274.0$50.0056/s

12Claude Sonnet 5.575.177.2$10.00107/s

13Claude Mythos 574.970.8$50.00—

14GPT-5.6 Sol74.773.0$22.0064/s

15GPT-6 Sol74.672.8$10.0061/s

16Claude Opus 574.573.8$25.0048/s

17Kimi K3.1 Pro74.475.7$10.0043/s

18GPT-5.4 Pro74.373.3$180.0021/s

19GPT-5.573.069.1$30.0052/s

20Muse Wonder71.773.1$20.0080/s

21dots3-note Preview71.363.4Free—

22GPT-5.6 Terra70.368.9$12.0066/s

23Hy4 preview70.266.4$2.5067/s

24Sage-A1.569.260.5——

24Ling 3.1 Flash69.267.2$0.13581/s

26Sakana Fugu-Ultra69.066.3——

27Gemini 3.7 Flash68.968.5$3.7585/s

28Claude Opus 4.868.667.7$25.0069/s

29MiniMax M3.1 Flash68.469.3$0.60100/s

30Muse Spark 1.368.170.1$4.2576/s

31MiMo-V2.6-Pro67.970.2$0.8731/s

32Gemini 3.8 Flash67.867.8$4.1368/s

32GPT-5.467.864.8$15.0038/s

34Kimi K366.966.9$16.0044/s

35Sakana Fugu66.163.9——

36DeepSeek V4 Pro 081365.865.5$4.4442/s

37Grok 4.665.767.8$6.0056/s

38Qwen3.8 Max65.665.6—41/s

38GPT-5.6 Luna65.665.5$1.2051/s

40GPT-5.3 Codex65.562.0$14.0052/s

41GPT-6 Luna65.465.5$0.5066/s

41Gemini 3.1 Pro65.462.3$12.00151/s

43DeepSeek V4 Flash Vision Exp65.360.4$1.3274/s

44Gemini 3.1 Pro Preview65.061.5$12.0077/s

45Claude Opus 4.7 (Adaptive)64.963.2$25.00—

45DeepSeek V4.1 Flash 101764.968.5$0.40261/s

47Muse Spark 1.264.664.9$4.25123/s

48Grok 4.764.369.1$6.0051/s

49Muse Spark Open64.265.1$4.2586/s

49Step 5 Preview64.268.4$2.70105/s

51Gemini 3.5 Flash64.164.2$9.00137/s

52Muse Spark 1.163.964.6$4.25152/s

53Grok 4.563.663.7$6.0053/s

53Gemini 3.6 Flash63.663.7$3.75121/s

53Claude Haiku 5.563.666.3$0.50159/s

56Inkling 2 Small63.466.4$1.00—

56Claude Sonnet 563.464.2$10.0075/s

58Qwen3.8 Max Preview63.064.2——

59Muse Spark62.659.8——

60Claude Opus 4.762.462.0$25.0047/s

61GPT-5.262.359.0$14.0045/s

61GLM-5.362.365.8$1.4086/s

63Qwen3.8 Max (0902)62.163.4$6.0033/s

64DeepSeek V4 Flash 073161.961.9$0.3061/s

65GPT-5.2-Codex61.259.9$14.0044/s

66Qwen3.8-Omni-Flash60.862.7$0.4741/s

67Grok 4.20 (Reasoning)60.755.5——

68Gemini 3 Pro60.557.9$12.00109/s

69Qwen3.7 Max60.460.5$4.4368/s

70Claude Opus 4.6 (Adaptive)60.360.0——

70Claude Opus 4.660.358.0$25.0036/s

72Beam 1.0 501B-A23B FP860.159.2——

72MiMo-V2.6-Flash60.164.3$0.2828/s

74Beam 1.0 501B-A23B NVFP459.158.5——

75Qwen3.8-Flash-Next58.964.3—55/s

76Beam 1.0 501B-A23B58.460.2—345/s

77Gemini 3 Flash Preview58.257.5$3.0078/s

78Mercury 358.052.3$0.35295/s

79GLM-5.3-Flash57.963.1$0.078122/s

79GLM-5.257.960.6$6.80111/s

81DeepSeek V4 Pro 042357.457.6$1.7889/s

82GPT-5.157.055.3$10.0050/s

83Claude Opus 4.5 Thinking56.856.3——

83Kimi K2.7 Code56.857.7$3.5035/s

85Kimi K2.656.658.2$3.4181/s

86GPT-556.455.0$10.0076/s

87Gemini 3 Flash56.355.0$3.00198/s

87Grok 4.2056.352.8$2.5072/s

87Qwen3.8-27B56.359.2$1.88113/s

90Qwen3.7 Plus56.057.7$1.2851/s

91Claude Sonnet 4.655.957.0$15.0044/s

92Inkling55.854.8$4.0585/s

93Apodex 1.155.658.9——

94Kimi K2.5 Thinking55.452.6——

95Qwen 3.6 Max (preview)55.357.2$6.1675/s

96Mistral Large 454.860.7$2.10113/s

97Beam 1.054.658.8——

98Grok 4.354.553.3$2.5061/s

98Qwen3.5 Plus Thinking54.548.1——

100Nemotron 3.5 Super VL 120B A12B54.451.0——

100GLM-5 Thinking54.453.4——

102A.X K254.351.9——

102Inkling-Small54.355.1$1.2080/s

104Grok Build 0.154.252.7$2.00147/s

105Gemini 3.1 Flash-Lite54.151.9$1.50134/s

105Nemotron 3.5 Super VL 120B A12B NVFP454.150.8——

107Apodex 1.1 Mini53.755.7Free—

107Claude Sonnet 4.5 Thinking53.753.8——

109Solar Open 253.653.3——

109GPT-5.1-Codex-Max53.653.8$10.0055/s

109GPT-5.1-Codex53.652.9$10.0055/s

112Qwen3.6 Plus53.454.6$1.9544/s

113Kimi K2.5 (Reasoning)53.353.0$3.00—

113Ternary Bonsai 2 27B53.352.2$0.5021/s

113Qwen3.5 397B (Reasoning)53.355.3$3.60—

116GPT-5.4 mini53.254.5$4.5095/s

116GPT-5 (high)53.253.6$10.0076/s

116Ripple-65B Preview53.252.3$0.6063/s

119Qwen3.5 397B53.153.8$3.6059/s

119MiMo-V2.5-Pro53.156.3$1.8047/s

121Kimi K2.553.052.2$3.3236/s

122MAI-Thinking-152.853.5——

123Claude Opus 4.552.652.0$25.0039/s

124GLM-5.152.556.0$4.40102/s

125Solar Pro 452.252.1$0.6044/s

126Gemini 3.1 Flash Lite Preview52.149.6$1.50151/s

127o351.951.8$8.0065/s

128Muse Glimmer 30B51.853.0$1.2090/s

128Hy3 Preview51.852.3$0.6069/s

130Qwen3.5-122B-A10B51.651.9$2.0870/s

130MiMo-V2.551.654.8$2.004/s

130Nemotron 3.5 Bolt 74B-A9B51.646.4——

133GPT-5 (medium)51.553.0$10.0076/s

134GPT-5.4 nano51.450.9$1.2591/s

135Grok 4.1 Fast (Reasoning)51.347.3——

136Grok 451.250.0—78/s

137DeepSeek V3.2 (Thinking)51.149.4$2.19—

138MiniMax M351.055.7$1.2079/s

139GLM-550.851.6$3.2045/s

140MiniMax M2.550.750.2$1.0844/s

141Muse Ember50.364.8$0.50158/s

142Nemotron 3 Ultra50.050.8Free—

143Grok 4 Fast (Reasoning)49.945.3——

143Kimi K2 Thinking49.948.2$2.5048/s

145Ling 3.0 Flash49.850.9$0.06373/s

146Qwen3.5-27B49.751.3$2.6069/s

147Qwen3.6-35B-A3B49.650.6$0.9065/s

147GLM-4.749.650.0$2.2028/s

149Qwen3.6-27B49.452.1$3.2060/s

150MiniMax M2.749.251.2$0.8435/s

151GLM-5-Turbo49.151.9$4.0025/s

151Gemini 2.5 Pro49.146.6$10.0091/s

153Qwen3 235B 2507 (Reasoning)49.047.4$3.50108/s

153Claude Opus 4.149.047.5$75.0012/s

155MiMo-V2-Pro48.951.8——

156Claude 4 Opus Thinking48.647.0——

157Qwen3.5-35B-A3B48.549.4$1.0043/s

157Gemma 4 26B A4B48.544.5Free—

157GPT-5 mini48.548.4$2.0078/s

160Step 3.7 Flash48.255.3$1.15112/s

161Gemini 3.5 Flash-Lite48.052.0$2.50130/s

161Claude 4.1 Opus Thinking48.047.0—29/s

163MiniMax M2.147.849.1$1.2034/s

164Claude Haiku 4.5 Thinking47.643.5——

164Gemma 4 31B47.647.9Free35/s

166Spark-X2.6-4B47.548.7——

167Claude Sonnet 4.547.147.9$15.0045/s

167Claude Sonnet 4 Thinking47.146.9——

169MiMo-V2-Omni46.947.2——

170GLM-5V-Turbo46.848.1$4.0030/s

170GLM-4.546.845.7$2.2025/s

172DeepSeek V3.1 (Reasoning)45.743.0——

173K-EXAONE 2.045.648.8——

174o4-mini45.444.7$4.40122/s

175Mercury 2.545.046.2$0.15125/s

176Qwen3 Max44.945.0$3.9062/s

177o144.844.5$60.0097/s

178GPT-OSS 120B44.743.4$0.6034/s

179DeepSeek V3.244.344.2$0.3819/s

180Mistral Medium 3.5 128B43.550.3$7.50—

180Kimi K2 Instruct43.543.2——

182DeepSeek-R1-052843.444.6$2.1520/s

183Grok 3 Mini43.244.3$0.5065/s

184Gemini 2.5 Flash (Preview)43.146.6——

185MiMo-V2-Flash42.843.4—129/s

185o3-mini42.840.0$4.40173/s

187DeepSeek-R142.338.9$2.5018/s

188LFM2.5-24B-A2B42.139.8$0.70280/s

188Qwen3.5-4B42.144.3—20/s

188Claude Opus 442.143.9——

191GPT-5 nano42.041.9$0.4081/s

192ZAYA1-8B41.938.2——

192K-Exaone41.942.3——

194Qwen3 235B 250741.442.3$0.8048/s

194Gemma 4 12B41.439.0——

196Trinity-Large-Thinking40.743.5$0.80138/s

196Gemini 2.5 Flash40.735.0$2.5095/s

196Claude Sonnet 440.743.0$15.0049/s

199Granite 4.2 30B40.541.6——

200Grok 3 [Beta]39.939.3——

201Exaone 4.0 32B39.832.4——

202Claude 3.7 Sonnet Thinking39.741.4——

202Mistral Small 439.737.9$0.60114/s

204Grok Code Fast 139.641.2$1.50172/s

205GPT-4.539.538.8$150.00—

206Nemotron Ultra 253B39.335.9——

206Kimi K239.339.6$2.3037/s

208MiniCPM5-2B39.237.1——

208Nemotron 3.5 Lightning 30B A3B NVFP439.237.0——

208GPT-4.139.238.0$8.0074/s

211Nemotron 3 Nano Omni 30B A3B39.138.6—228/s

211GPT-OSS 20B39.141.7$0.13310/s

213GLM-4.638.341.8$2.0030/s

214Claude Haiku 4.538.243.2$5.0088/s

215Command A+38.040.2$1.50161/s

216Granite 4.2 8B37.737.6$0.15101/s

217Trinity-Large-Preview37.542.2$1.00—

218Mistral Medium 3.537.342.7$7.5092/s

219DeepSeek V3 032437.237.1——

220Sarvam 105B37.036.7—191/s

221North Mini Code36.938.0Free70/s

222Qwen3 235B-A22B36.738.1$1.8244/s

223Claude 3.7 Sonnet36.637.7——

224Llama4 Maverick Instruct Basic36.536.0——

224DeepSeek V3.136.537.9—98/s

226GPT-4.1 mini36.434.5$1.6072/s

227LFM3-1.7B36.241.5$0.20218/s

228LongCat-Flash-Lite-Sparse36.139.3——

229Nemotron 3 Nano 30B35.932.9—270/s

229Mistral Large 335.934.3$1.5061/s

231QwQ-32B35.336.3——

231Solar Pro 335.339.2$0.60114/s

231Celeris-135.333.7$0.701,555/s

234Magistral Medium35.133.7——

235o1-mini34.033.5$4.40—

236Magistral Small33.531.4——

237Llama 4 Maverick33.134.4$1.0061/s

238Gemini 2.0 Flash33.029.5——

239Grok 4.1 Fast Non Reasoning32.736.2——

239Grok 4.1 Fast32.733.3$0.50138/s

241Sarvam 30B32.430.3——

242Mistral Medium 332.027.1$2.0094/s

243DeepSeek V331.829.9$1.1037/s

243Ling 2.6 Flash31.837.0—104/s

245GPT-4o31.728.7$10.0066/s

246Claude 3.5 Sonnet31.332.2$15.00—

247Granite 4.2 3B30.933.2——

248Gemma 4 E4B30.528.2——

249Claude 4 Sonnet30.439.5$15.0040/s

250Phi-4-reasoning30.327.1—41/s

251GPT-4o mini29.626.2$0.6056/s

252Grok 229.529.1——

253Mistral Large 229.227.4$6.0070/s

253GPT-4.1 nano29.226.3$0.4056/s

255Gemini 2.5 Flash-Lite28.727.1$0.40140/s

256Nova Pro28.425.6—129/s

257Llama 4 Scout28.226.8$0.3063/s

258LFM2.5-2.6B26.929.3$0.50189/s

259Claude 3 Opus26.423.3$75.00—

260Devstral Small26.328.2——

261Gemma 3 27B26.025.3—31/s

262Llama 3.1 405B25.327.0—29/s

263Phi-425.222.4$0.1493/s

264LFM2.5-8B-A1B25.130.6—335/s

265Command A24.823.9——

266Mistral Small 3.124.024.2—28/s

266Llama 3.3 70B24.028.2——

266Claude 3.5 Haiku24.024.1——

269Gemma 4 E2B23.821.9——

270Nemotron 3.5 Lightning 30B A3B22.628.1——

271Llama 3.1 70B21.724.8——

272Claude 3 Haiku21.322.0$1.2573/s

273Gemini 1.5 Flash20.120.2——

274Granite-4.0-H-1B17.816.3——

275LLaDA2.2-mini16.824.3——

275Granite-4.0-H-350M16.812.7——

277MiniCPM5-1B15.615.5——

278Granite-4.0-350M14.111.9——

279Llama 3.1 8B12.217.9——