fedi-index：基於開放資料的 AI 模型排名

我們自己的語言模型排名：人類偏好投票與獨立基準測試結果放在同一尺度上，每個分數都附有不確定性與分項拆解。

2026-10 即時資料（每日重新計算） · 方法論 v1 · 方法論 · 歷史 · JSON

全部開放權重封閉

繁體中文最佳模型暫定查看全部

等級

模型

fedi-index

百分位

人類偏好

解題能力

性價比

來源

1,504 ±8

100.0

1,496

1,520

+76◆

›

1,498 ±8

99.7

1,490

1,512

+47

›

1,493 ±7

1,492

1,494

+59

›

1,490 ±5

1,509

—

›

1,484 ±7

1,479

1,492

+33

›

1,476 ±5

1,492

1,452

+41

›

1,472 ±7

1,477

1,464

+84◆

›

1,471 ±22

—

1,499

-10

›

1,469 ±8

1,470

1,468

+74

›

1,469 ±5

1,476

1,460

+35

›

1,469 ±7

1,454

1,493

+40

›

1,469 ±6

1,467

1,471

+81

›

1,466 ±6

1,460

1,476

+29

›

1,466 ±6

1,466

1,465

+44

›

1,465 ±9

1,437

1,513

+53

›

1,463 ±8

1,437

1,514

+12

›

1,462 ±6

1,460

1,467

+50

›

1,461 ±6

1,454

1,470

+27

›

1,459 ±5

1,465

1,451

+54

›

1,459 ±5

1,457

1,462

+39

›

1,457 ±6

1,458

1,456

+42

›

1,457 ±5

1,460

1,453

+49

›

1,457 ±7

1,459

1,455

+62

›

1,457 ±6

1,460

1,453

+60

›

1,456 ±5

1,474

—

+93◆

›

1,456 ±5

1,458

1,453

+68

›

1,453 ±7

1,440

1,476

+38

›

1,452 ±7

1,455

1,449

+58

›

1,449 ±7

1,453

1,444

+39

›

1,449 ±15

—

1,473

-32

›

1,448 ±10

1,452

1,441

+92◆

›

1,448 ±6

1,454

1,440

+33

›

1,448 ±6

1,459

1,432

+105◆

›

1,443 ±6

1,438

1,452

+32

›

1,443 ±6

1,453

1,428

+21

›

1,443 ±2

1,457

—

+80

›

1,442 ±5

1,453

1,427

+45

›

1,440 ±5

1,436

1,446

+35

›

1,440 ±2

1,454

—

›

1,439 ±8

1,445

1,430

—

›

1,438 ±6

1,426

1,457

+80

›

1,438 ±17

—

1,457

-39

›

1,437 ±4

1,440

1,434

+57

›

1,437 ±6

1,422

1,460

+32

›

1,437 ±2

1,451

—

+52

›

1,436 ±6

1,448

1,417

›

1,435 ±11

—

1,454

+40

›

1,435 ±5

1,443

1,423

+44

›

1,434 ±8

1,397

1,494

+22

›

1,433 ±4

1,449

—

+98◆

›

1,432 ±5

1,445

1,415

+36

›

1,430 ±25

—

1,457

+14

›

1,428 ±5

1,421

1,438

+12

›

1,427 ±5

1,437

1,415

+33

›

1,426 ±4

1,432

1,418

+18

›

1,425 ±8

1,419

1,432

+33

›

1,425 ±7

1,432

1,416

+46

›

1,424 ±13

—

1,445

+89

›

1,424 ±7

1,432

1,414

+20

›

1,421 ±5

1,436

1,402

+52

›

1,420 ±5

1,438

1,395

-2

›

1,419 ±7

1,395

1,455

+80

›

1,419 ±6

1,434

—

+32

›

1,419 ±2

1,432

—

+31

›

1,418 ±6

1,426

1,407

+26

›

1,418 ±5

1,430

1,405

+36

›

1,416 ±6

1,393

1,444

+11

›

1,415 ±5

1,423

1,404

+34

›

1,415 ±5

1,430

—

›

1,414 ±2

1,426

1,374

+26

›

1,413 ±2

1,426

—

+24

›

1,413 ±5

1,433

1,387

›

1,413 ±8

1,424

1,392

+50

›

1,412 ±5

1,425

1,396

+27

›

1,412 ±6

1,428

1,390

+23

›

1,412 ±2

1,424

—

+31

›

1,411 ±3

1,425

—

+33

›

1,411 ±3

1,424

—

+68

›

1,411 ±2

1,423

—

+75

›

1,407 ±3

1,420

—

+14

›

1,407 ±7

1,419

1,392

+71

›

1,407 ±6

1,403

1,413

+38

›

1,406 ±2

1,419

—

›

1,406 ±15

—

1,422

-63

›

1,406 ±3

1,419

—

+46

›

1,405 ±5

1,403

1,408

+14

›

1,404 ±6

1,395

1,417

-4

›

1,404 ±2

1,416

—

+26

›

1,403 ±6

1,418

1,384

+29

›

1,403 ±5

1,411

1,394

+54

›

1,402 ±2

1,414

—

›

1,400 ±6

1,418

1,375

-61

›

1,400 ±6

1,423

1,365

›

1,400 ±11

—

1,416

›

1,400 ±2

1,411

—

-12

›

1,399 ±3

1,405

1,374

-5

›

1,398 ±7

1,426

1,359

+58

›

1,398 ±6

1,416

1,373

+20

›

1,398 ±3

1,410

—

-24

›

100

1,396 ±3

1,408

—

+18

›

101

1,396 ±2

1,408

—

+25

›

102

1,395 ±8

1,406

1,383

+46

›

103

1,395 ±6

1,395

-27

›

104

1,394 ±4

1,407

—

+27

›

105

1,394 ±10

1,402

1,375

+32

›

106

1,393 ±4

1,380

1,411

›

107

1,392 ±5

1,386

1,400

-17

›

108

1,391 ±8

1,416

1,358

+54

›

109

1,391 ±1

1,403

—

+19

›

110

1,391 ±2

1,402

—

+14

›

111

1,390 ±6

1,403

—

+31

›

112

1,389 ±5

1,398

1,378

+12

›

113

1,388 ±3

1,400

—

+32

›

114

1,388 ±5

1,400

—

›

115

1,387 ±14

—

1,401

-78

›

116

1,387 ±2

1,398

—

+55◆

›

117

1,386 ±8

1,409

1,356

-19

›

118

1,386 ±2

1,397

—

+14

›

119

1,383 ±6

1,400

1,358

-12

›

120

1,382 ±9

1,393

1,364

›

121

1,381 ±2

1,392

—

+50

›

122

1,381 ±11

—

1,395

›

123

1,380 ±6

1,403

1,353

›

124

1,380 ±2

1,391

—

+16

›

125

1,380 ±3

1,391

—

+18

›

126

1,380 ±7

1,393

—

›

127

1,379 ±4

1,391

—

›

128

1,378 ±6

1,396

1,356

›

129

1,378 ±14

—

1,391

-2

›

130

1,377 ±10

1,399

1,340

+19

›

131

1,376 ±7

1,383

1,365

+11

›

132

1,374 ±6

1,381

1,365

-86

›

133

1,374 ±4

1,386

—

-17

›

134

1,374 ±2

1,384

—

+38

›

135

1,373 ±6

1,367

1,383

+14

›

136

1,372 ±2

1,382

—

-45

›

137

1,371 ±5

1,362

1,382

›

138

1,371 ±9

1,357

1,392

›

139

1,370 ±3

1,381

—

+47◆

›

140

1,370 ±5

1,382

—

›

141

1,366 ±12

—

1,380

›

142

1,365 ±5

1,376

—

›

143

1,364 ±15

—

1,379

+57◆

›

144

1,360 ±15

—

1,375

›

145

1,359 ±6

1,359

-62

›

146

1,356 ±6

1,338

1,382

-37

›

147

1,356 ±9

1,375

1,320

+25

›

148

1,356 ±7

1,362

1,349

›

149

1,356 ±4

1,366

—

›

150

1,356 ±9

1,363

1,346

-22

›

151

1,356 ±2

1,365

—

-2

›

152

1,355 ±2

1,364

—

+19

›

153

1,353 ±12

—

1,366

-4

›

154

1,353 ±6

1,370

1,324

-56

›

155

1,351 ±2

1,360

—

+16

›

156

1,348 ±4

1,358

—

›

157

1,347 ±6

1,357

—

-12

›

158

1,346 ±3

1,356

—

-10

›

159

1,345 ±6

1,355

—

›

160

1,344 ±3

1,353

—

-22

›

161

1,344 ±8

1,345

1,341

-2

›

162

1,341 ±4

1,353

1,283

-31

›

163

1,338 ±3

1,347

—

-36

›

164

1,334 ±4

1,343

—

›

165

1,334 ±6

1,344

—

-20

›

166

1,334 ±13

—

1,370

-74

›

167

1,332 ±4

1,342

1,286

›

168

1,332 ±4

1,340

—

-17

›

169

1,331 ±2

1,339

—

›

170

1,330 ±16

—

1,342

›

171

1,330 ±4

1,338

—

-32

›

172

1,329 ±17

—

1,341

-39

›

173

1,328 ±5

1,337

—

›

174

1,328 ±17

—

1,338

-10

›

175

1,327 ±6

1,337

1,311

-42

›

176

1,325 ±2

1,333

—

-30

›

177

1,324 ±3

1,332

—

›

178

1,323 ±8

1,316

1,334

-7

›

179

1,321 ±2

1,329

—

-36

›

180

1,319 ±3

1,326

—

›

181

1,318 ±4

1,326

—

›

182

1,318 ±4

1,326

—

›

183

1,317 ±3

1,325

—

-14

›

184

1,316 ±5

1,323

—

-60

›

185

1,315 ±6

1,323

—

-40

›

186

1,310 ±6

1,318

—

-14

›

187

1,309 ±9

1,309

1,310

-29

›

188

1,304 ±3

1,310

—

›

189

1,303 ±7

1,310

—

›

190

1,296 ±8

1,302

1,285

—

›

191

1,295 ±8

1,303

—

›

192

1,295 ±5

1,302

—

-134

›

193

1,295 ±18

—

1,302

-78

›

194

1,294 ±7

1,301

—

›

195

1,293 ±9

1,273

1,325

-26

›

196

1,292 ±3

1,297

—

›

197

1,289 ±4

1,294

—

›

198

1,288 ±6

1,295

—

›

199

1,287 ±5

1,294

—

›

200

1,284 ±5

1,289

—

›

201

1,283 ±3

1,288

—

›

202

1,282 ±5

1,288

—

-28

›

203

1,279 ±3

1,284

—

›

204

1,279 ±9

1,282

1,273

-57

›

205

1,276 ±13

1,283

—

›

206

1,270 ±8

1,275

—

›

207

1,268 ±9

1,270

1,265

—

›

208

1,266 ±5

1,270

—

-95

›

209

1,259 ±3

1,262

—

›

210

1,258 ±5

1,262

—

›

211

1,250 ±3

1,253

—

›

212

1,249 ±3

1,252

—

›

213

1,241 ±6

1,244

—

›

214

1,241 ±5

1,243

—

›

215

1,230 ±4

1,232

—

-178

›

216

1,227 ±3

1,228

—

›

217

1,222 ±4

1,223

—

›

218

1,222 ±6

1,223

—

›

219

1,219 ±3

1,220

—

›

220

1,217 ±7

1,218

—

›

221

1,216 ±4

1,221

1,196

—

›

222

1,216 ±4

1,217

—

›

223

1,215 ±7

1,216

—

›

224

1,207 ±4

1,207

—

›

225

1,207 ±3

1,207

—

›

226

1,199 ±5

1,198

—

-133

›

227

1,198 ±2

1,198

—

-220

›

228

1,194 ±7

1,193

—

›

229

1,184 ±4

1,183

—

-160

›

230

1,183 ±3

1,182

—

›

231

1,178 ±3

1,176

—

›

232

1,174 ±4

1,172

1,181

—

›

233

1,173 ±4

1,171

—

›

234

1,169 ±3

1,167

—

-236

›

235

1,168 ±4

1,166

—

›

236

1,163 ±7

1,160

—

›

237

1,160 ±4

1,153

1,188

—

›

238

1,151 ±7

1,148

—

›

239

1,149 ±3

1,146

—

-199

›

240

1,148 ±6

1,144

—

›

241

1,147 ±4

1,144

—

›

242

1,142 ±4

1,138

—

›

243

1,142 ±4

1,138

—

›

244

1,140 ±7

1,135

—

-275

›

245

1,139 ±3

1,135

—

›

246

1,136 ±6

1,131

—

›

247

1,135 ±4

1,131

—

›

248

1,128 ±6

1,122

—

›

249

1,126 ±3

1,121

—

-219

›

250

1,124 ±5

1,119

—

›

251

1,121 ±4

1,116

—

›

252

1,116 ±5

1,110

—

›

253

1,111 ±7

1,104

—

›

254

1,110 ±6

1,104

—

›

255

1,110 ±4

1,103

—

›

256

1,109 ±3

1,103

—

›

257

1,107 ±4

1,101

—

›

258

1,107 ±5

1,100

—

›

259

1,106 ±3

1,099

—

›

260

1,105 ±4

1,099

—

-230

›

261

1,105 ±6

1,098

—

›

262

1,105 ±8

1,097

—

›

263

1,101 ±10

1,092

—

›

264

1,101 ±8

1,093

—

›

265

1,099 ±4

1,092

—

-243

›

266

1,096 ±9

1,087

—

›

267

1,089 ±11

1,078

—

›

268

1,086 ±6

1,077

—

›

269

1,083 ±12

1,073

—

›

270

1,082 ±5

1,074

—

›

271

1,082 ±8

1,072

—

›

272

1,082 ±10

1,071

—

›

273

1,080 ±5

1,071

—

›

274

1,078 ±4

1,070

—

›

275

1,070 ±4

1,061

—

-272

›

276

1,067 ±8

1,056

—

›

277

1,059 ±6

1,048

—

›

278

1,056 ±6

1,046

—

›

279

1,054 ±6

1,043

—

-274

›

280

1,052 ±6

1,041

—

›

281

1,051 ±7

1,040

—

›

282

1,051 ±4

1,041

—

›

283

1,038 ±5

1,026

—

›

284

1,014 ±7

1,000

—

›

285

1,010 ±6

996

—

›

286

1,005 ±10

988

—

›

287

984 ±8

967

—

›

288

977 ±7

960

—

›

289

969 ±8

951

—

›

290

966 ±6

945

—

›

291

961 ±7

943

—

›

292

931 ±8

910

—

›

293

904 ±8

880

—

›

AArena（人類投票）EEpoch AI（基準測試） ◆ 帕累托前緣：沒有模型能同時更好又更便宜

fedi-index 是固定尺度上的 Elo 等值：在同一方法論版本內，各月份可互相比較。等級 S–D：依分項內的百分位決定，且不會高於信賴區間所允許的等級（門檻見方法論）。「暫定」——此處只有一個來源系列測量過該模型。方法論

性價比：以價格衡量品質

每個點代表一個付費 API 模型：fedi-index 對照混合價格（輸入 3：輸出 1，各供應商中位數）。線條為帕累托前緣——品質與價格都無人能同時超越的模型。

fedi-index 與價格帕累托前緣：沒有模型能同時更好又更便宜 ↑ fedi-index · → 每 100 萬 token 的美元價格（對數）

帕累托前緣：9 個模型

Qwen3.7 Flash 1,364 · US$0.055
Solar Pro 4 1,370 · US$0.105
MiMo-V2-Flash 1,387 · US$0.15
MiMo-V2.6-Flash 1,433 · US$0.175
GLM 5.3 Flash 1,448 · US$0.2375
DeepSeek V4.1 Flash 1,448 · US$0.42
MiMo-V2.6-Pro 1,456 · US$0.5438
Gemini 3.8 Flash 1,472 · US$1.50
Claude Opus 5.5 1,504 · US$8

在你的硬體上

開放模型在常見家用配置上的排名：綜合分數，若生成速度低於舒適水準或量化低於 Q4 則扣分。

GPU 12 GB

1 Qwen3 Next 80B A3B Instruct Q5_K_M · 部分放在 RAM ≈24 tok/s 62
2 Gemma 4 26B A4B UD-Q8_K_XL · 部分放在 RAM ≈15 tok/s 54
3 Qwen3.5-35B-A3B Q8_0 · 部分放在 RAM ≈18 tok/s 52
4 Qwen3.6 35B A3B UD-Q8_K_XL · 部分放在 RAM ≈17 tok/s 47
5 Qwen3 Next 80B A3B Thinking Q5_K_M · 部分放在 RAM ≈24 tok/s 46
6 Qwen3.5-122B-A10B UD-IQ4_NL · 部分放在 RAM ≈10 tok/s 45
7 Qwen3 30B A3B Instruct 2507 Q8_0 · 部分放在 RAM ≈16 tok/s 45
8 gpt-oss-120b MXFP4 · 部分放在 RAM ≈19 tok/s 44
9 Qwen3.5-9B Q6_K ≈36 tok/s 42
10 Qwen3.8 27B UD-Q4_K_XL · 部分放在 RAM ≈4 tok/s 37

GPU 24 GB

1 Qwen3.8 27B UD-Q5_K_S ≈38 tok/s 81
2 Gemma 4 31B Q4_1 ≈37 tok/s 67
3 Gemma 4 26B A4B UD-Q5_K_S ≈129 tok/s 63
4 Qwen3 Next 80B A3B Instruct UD-Q6_K_XL · 部分放在 RAM ≈26 tok/s 62
5 Qwen3.5-27B Q5_K_S ≈38 tok/s 60
6 Qwen3.6 27B Q5_K_S ≈37 tok/s 59
7 Qwen3.5-35B-A3B Q4_K_S ≈191 tok/s 56
8 Qwen3.6 35B A3B UD-Q4_K_S ≈189 tok/s 51
9 Qwen3 30B A3B Instruct 2507 Q5_K_S ≈147 tok/s 50
10 Qwen3.5-122B-A10B Q4_K_S · 部分放在 RAM ≈11 tok/s 47

Mac 64 GB

1 Qwen3.8 27B Q8_0 ≈13 tok/s 64
2 Gemma 4 26B A4B UD-Q8_K_XL ≈40 tok/s 63
3 Qwen3 Next 80B A3B Instruct Q4_K_S ≈84 tok/s 62
4 Qwen3.5-122B-A10B UD-IQ3_XXS ≈44 tok/s 56
5 Qwen3.5-35B-A3B Q8_0 ≈51 tok/s 56
6 Qwen3.6 35B A3B UD-Q8_K_XL ≈49 tok/s 51
7 Gemma 4 31B Q8_0 ≈11 tok/s 50
8 Qwen3 30B A3B Instruct 2507 Q8_0 ≈45 tok/s 50
9 Qwen3.5-27B Q8_0 ≈13 tok/s 48
10 Qwen3.6 27B Q8_0 ≈13 tok/s 47

128 GB 統一記憶體

1 DeepSeek V4 Flash 0423 UD-IQ3_XXS ≈26 tok/s 65
2 Inkling Small UD-IQ3_S ≈23 tok/s 65
3 MiniMax M2.7 UD-IQ4_NL ≈20 tok/s 64
4 Gemma 4 26B A4B UD-Q8_K_XL ≈29 tok/s 63
5 Qwen3 Next 80B A3B Instruct Q8_0 ≈36 tok/s 62
6 Qwen3.8 27B UD-Q4_K_M ≈10 tok/s 58
7 Qwen3.5-35B-A3B Q8_0 ≈37 tok/s 56
8 Step 3.5 Flash IQ4_XS ≈17 tok/s 55
9 Qwen3.5-122B-A10B Q6_K ≈15 tok/s 54
10 DeepSeek V4 Flash 0731 UD-IQ3_XXS ≈11 tok/s 53

CPU，32 GB RAM

1 Gemma 4 26B A4B UD-Q4_K_XL ≈6 tok/s 36
2 Qwen3.5-35B-A3B Q4_K_M ≈7 tok/s 32
3 Qwen3.6 35B A3B UD-Q4_K_XL ≈7 tok/s 30
4 Qwen3 30B A3B Instruct 2507 Q4_K_M ≈6 tok/s 28
5 Qwen3.8 27B UD-Q4_K_XL ≈2 tok/s 27
6 GLM 4.7 Flash Q4_K_M ≈6 tok/s 24
7 Nemotron 3 Nano 30B A3B Q4_K_M ≈7 tok/s 23
8 Gemma 4 31B Q4_K_M ≈2 tok/s 22
9 Qwen3 30B A3B Q4_K_M ≈6 tok/s 21
10 Qwen3.5-27B Q4_K_M ≈2 tok/s 21

能否載入與速度為我們的估算（權重 + KV 快取對照記憶體，速度依頻寬推算）。檢查你自己的配置：能在本機執行嗎？

嵌入你的網站

將表格作為小工具嵌入你的網站：自動更新，並保留來源標示。

程式碼

<script src="https://fedi.software/embed.js" data-fig="fedi-index" data-lang="cn" async></script>