88Z.ai / RANKINGS

AI model speed and latency leaderboard

Based on Artificial Analysis time-per-task and first-token latency; lower times rank higher.

Last collected by this site: 2026-11-15 06:40 (UTC+8, Asia/Shanghai). Source publication dates are in the table or the citations below.

Based on Artificial Analysis time-per-task and first-token latency; lower times rank higher.

202 model families (best published variant per family)

Click a column header to sort. The table is fully server-rendered and remains readable without JavaScript.

#ModelTime / task (min)First token (s)AA Index
1Gemini 4 Carbon (High)$5.481—31.3
2Gemini 4 Argon (High)$4.971—29.6
3Muse Wonder (Max)$6.951—14.3
4Gemini 3.8 Flash (High)$7.771—14.6
5Gemini 3.1 Pro Preview--1——
6GLM 5.3 (Max)$5.011—25
7GPT 6.1 Sol (Max)$1.271—23.3
8Mimo v2.6 Pro小米集团1—35.7
9GLM 5.3 FlashZhipu1—45.5
10GPT 6 Astra (Max)$6.241—22.8
11qwen3.7 Plus--1——
12Gemini 3.5 Flash Lite$0.691—1.7
13Step 5 Preview阶跃星辰1—14.3
14Mimo v2.6 Flash小米集团1—50.3
15GPT 6 Luna (Max)$0.151—12.2
16MiniMax M3稀宇科技1—13.5
17Inkling 2 Small$1.181—30.9
18Inkling Small$0.201—5.5
19Muse Wonder Full (Max)--1——
20GPT 6.1 Terra (Max)$0.961—14.3
21Muse Ember (Max)$0.141—3.3
22MiniMax m3.1 Flash稀宇科技1—16.6
23Gemini 4 Hydrogen (xHigh)$3.161—14.7
24Kimi k3.1 ProMoonshot1—39.7
25Claude Haiku 5.5 (Max)$0.411—14.8
26Tencent Hy4$1.031—14.2
27DeepSeek v4.1 Flash 1017 (Max)$0.601—13.9
28Ling 3.1 Flash 1019$3.191——
29Mistral Large 4 (High)$2.061—13.4
30Muse Spark Open (xHigh)$2.021—7.6
31Agnes 2.5 Pro Theta$0.571—24.9
32DeepSeek V4 Pro 0813 (Max)$1.771—19.2
33Beam 1.0$0.281—2.1
34Nemotron 3.5 Super Vl$0.561—5.8
35Inkling (xHigh)$1.351—10.6
36Nemotron 3.5 Lightning$0.311—4.1
37Nova 2.0 Omni (Medium)--1——
38Kimi Linear 48b A3b Instruct--1——
39Longcat 2.0--1——
40Nova 2.0 Lite (High)--1——
41Solar Mini 4--1——
42DeepSeek V4 Flash Vision (Max)--1——
43Solar Open2 250b--1——
44Llama 4 MaverickMeta1——
45GPT 6 Sol Daybreak Blue (Max)--1——
46Nemotron Cascade 2 30b A3b--1——
47Ling 3.1 Flash--1——
48Nova Premier--1——
49Agnes 2.5 Pro Beta--1——
50Claude Fable 5.2 Max Default Fallback$26.541—44.3
51Claude Opus 5.5 Max Default Fallback$10.041—25.2
52Claude Sonnet 5.5 Max Default Fallback$10.021—33.6
53Nvidia Nemotron 3 Nano 30b A3b Non Reasoning--1——
54Quasar 438b Max Based On GLM 5.2--1——
55Molmo 7b D--4——
56Tiny Aya Global--8——
57Llama 4 Scout--10——
58Phi 4--16——
59Gemma 3 270m--32——
60lfm2.5 Vl 1.6b--32——
61lfm2.5 1.2b Instruct--32——
62lfm2.5 1.2b Thinking--32——
63Tri 21b Think--32——
64Deephermes 3 Mistral 24b Preview Non Reasoning--32——
65Granite 4.0 350m--33——
66Granite 4.0 H 350m--33——
67lfm2.5 8b A1b--33——
68Lfm2 24b A2b--33——
69Molmo2 8b--37——
70Exaone 4.0 1.2b Non Reasoning--64——
71Olmo 3.1 32b Instruct--66——
72Olmo 3.1 32b Think--66——
73Olmo 3 7b Think--66——
74Apertus 70b Instruct--66——
75Sarvam 30b (High)--66——
76Apertus 8b Instruct--66——
77Step3 Vl 10b--66——
78Qwen3 Omni 30b A3b Instruct--66——
79Olmo 3 7b Instruct--66——
80Qwen3 Omni 30b A3b Reasoning--66——
81Ring Flash 2.0--128——
82Granite 4.0 H Small--128——
83Granite 4.0 Micro--128——
84Hyperclova X Seed Think 32b--128——
85Jt (Mini)--128——
86Sarvam 105b (High)--128——
87Gemma 4 E4b Non Reasoning--128——
88Magistral Small 1.2--128——
89Gemma 4 E2b Non Reasoning--128——
90Magistral Medium 1.2--128——
91R1 1776--128——
92Gemini 3 Deep Think--128——
93Ernie 5.0 Thinking Preview--128——
94Apriel v1.6 15b Thinker--128——
95Granite 4.0 H 1b--128——
96Reka Flash 3--128——
97Mi Dm K 2.5 Pro--128——
98lfm2.5 2.6b--128——
99Hermes 4 Llama 3.1 70b Non Reasoning--128——
100Phi 4 Multimodal Instruct--128——
101Llama 3.1 Instruct 405b--128——
102Solar Open 100b Reasoning--128——
103Llama 3.2 Instruct 11b Vision--128——
104Motif 2 12.7b Reasoning--128——
105Nvidia Nemotron Nano 12b V2 Vl Reasoning--128——
106Minicpm5 1b Reasoning--128——
107Hermes 4 Llama 3.1 405b Reasoning--128——
108Llama 3.3 Instruct 70b--128——
109Llama 3.2 Instruct 90b Vision--128——
110Phi 4 Mini Instruct--128——
111Deephermes 3 Llama 3.1 8b Preview Non Reasoning--128——
112Llama 3.1 Nemotron Ultra 253b V1 Reasoning--128——
113Cogito v2.1 Reasoning--128——
114Nova Micro--130——
115Granite 4.2 30b--131——
116Granite 4.2 8b$0.0405131—12
117Granite 4.2 3b--131——
118Muse Glimmer (High)$0.18131—5.2
119GPT Oss 120b (High)$0.35131—7.5
120GPT Oss 20b (Low)--131——
121Nvidia Nemotron Nano 9b V2 Non Reasoning--131——
122Exaone 4.0 32b Non Reasoning--131——
123G9v3 3b--131——
124Minicpm5 2b--131——
125Celeris 1--131——
126K2 Horizon 0.9b--131——
127Ernie 4.5 300b A47b--131——
128G9v3 39a5b--131——
129Hypernova 60b 2605 High Based On GPT Oss 120b--131——
130Intellect 3 Based On GLM 4.5 Air--131——
131Nemotron 3.5 Bolt E4b 4gb Reasoning--131——
132Command A+$0.00192—3.2
133O3OpenAI200——
134Hy3--256——
135Mistral Medium 3.5$1.04256—4.5
136qwen3.8 Flash Next$0.57256—33.8
137Kimi k2.7 Code--256——
138qwen3.8 27b (xHigh)$3.53256—42.7
139Apodex 1.1$0.76256—14.8
140North Mini Code$0.00256—2.9
141Jt 35b Flash--256——
142qwen3.5 Omni Plus--256——
143Nova 2.0 Pro Preview (Medium)--256——
144Command A--256——
145Doubao Seed Code--256——
146Longcat Flash Lite--256——
147Ministral 3 3b--256——
148qwen3.5 Omni Flash--256——
149Ministral 3 14b--256——
150Jamba 1.7 Large--256——
151nanbeige4.1 3b--256——
152Ministral 3 8b--256——
153Diffusiongemma 26b A4b--256——
154Falcon H1r 7b--256——
155Kat Coder Pro V2--256——
156Qwen3 Coder Next--256——
157Gemma 4 31b Reasoning--256——
158Mistral Small 4 Reasoning$0.0399256—1.7
159Gemma 4 26b A4b Reasoning--256——
160Nemotron 3 Nano Omni 30b A3b Reasoning--256——
161Jamba 1.7 (Mini)--258——
162Mercury 2.5--260——
163Qwen3 Next 80b A3b Instruct--262——
164Jt 4.1 Flash 236b A21b--262——
165K Exaone 2.0--262——
166qwen3.5 122b A10b Non Reasoning--262——
167Gemma 4 12b Non Reasoning--262——
168qwen3.5 2b Non Reasoning--262——
169qwen3.5 9b Non Reasoning--262——
170qwen3.5 4b Non Reasoning--262——
171Exaone 4.5 33b Non Reasoning--262——
172qwen3.5 0.8b Non Reasoning--262——
173Ling 3.0 Flash Vl--262——
174Minicpm V 4.6 1.3b--262——
175K2 Think V2--262——
176qwen3.6 35b A3b Non Reasoning--262——
177Ling 3.0 Flash Fin--262——
178Nemotron 3.5 Bolt 74b A9b--262——
179Ling 3.0 Tiny--262——
180Ax K2--262——
181qwen3.8 2.4t A95b--262——
182Motif 3--262——
183Ring 2.6 1t--262——
184Jamba Reasoning 3b--262——
185Nemotron 3 Ultra 550b A55b Reasoning$2.21262—4.3
186Nvidia Nemotron 3 Nano 4b--262——
187qwen3.5 397b A17b Reasoning--262——
188Nex N2 Pro Based On qwen3.5 397b A17b--262——
189Qwen3 Next 80b A3b Reasoning--262——
190GPT 5.3 Codex (xHigh)--400——
191GPT 5.5 Instant June 2026--400——
192Grok 4.7 (xHigh)$7.77500—26.6
193Solar Pro 4$0.96512—63.3
194Trinity Large Thinking--512——
195K2 V2 (High)--512——
196nanbeige4.3 33b--524——
197K2 Horizon 375b A23b$0.00524—4.4
198K2 Horizon 7b--524——
199K2 Horizon 3.7b--524——
200K2 Horizon Mova 36b A4b--524——
201GPT 5.5 Pro (xHigh)--922——
202qwen3.8 Max 0902阿里巴巴984—48.6

Sources

FAQ

Where do these scores come from?
Every score comes from a public benchmark or official machine-readable file, with a source link and date. This site does not invent scores, and missing values are not treated as zero.
How is the overall score calculated?
Each source is min-max normalized within itself, then weighted: Artificial Analysis 28%, LMArena Text Arena 28%, LiveBench 22%, coding composite 15%, reasoning composite 7%. Missing sources are dropped and the rest are renormalized. A model needs at least two headline sources (or three partials) to enter the overall board. When several variants of one family exist, only the best overall variant is shown.
Are site prices part of the benchmarks?
No. Benchmark scores come from third-party public boards. CNY prices are this site’s catalogue reference rates (CNY 1 = 1,000 points) and are settled on actual usage. They are shown separately so billing is not confused with quality.
How often is this updated?
A collector runs automatically every day. The upstream boards may not publish a new snapshot daily, so each page shows both this site’s collection time and the source publication date.