Output Speed (tokens/s)

Median output tokens per second measured by AA (median across providers, or first-party API).

211models

–Open Index weight

–last updated

Artificial Analysispublisher

#	Model	Effort	Score	Cost / task	Harness
1	Celeris-1	none	1380	–	–
2	Mercury 2.5	–	777	–	–
3	Mercury 2	high	542	–	–
4	Gemini 3.5 Flash-Lite	high	446	–	–
5	Muse Spark 1.2	xhigh	375	–	–
6	Gemini 3.1 Flash-Lite	high	368	–	–
7	Trinity Large Thinking	–	346	–	–
8	Gemini 2.5 Flash Lite	–	342	–	–
9	Muse Spark 1.3	xhigh	323	–	–
10	Nova Micro	none	311	–	–
11	Nemotron 3.5 Lightning	–	303	–	–
12	Gemini 3.7 Flash	low	302	–	–
13	gpt-oss-20b	low	283	–	–
14	Ling 3.0 Flash	–	278	–	–
15	GPT-5.4 Mini	xhigh	275	–	–
16	NVIDIA Nemotron 3 Nano 30B A3B (Non-reasoning)	none	253	–	–
17	Qwen 3.5	none	241	–	–
18	Nemotron 3 Nano Omni 30B A3B Reasoning	–	237	–	–
19	Haiku 5.5	max	237	–	–
20	NVIDIA Nemotron 3 Nano 30B A3B (Reasoning)	–	233	–	–
21	Nemotron 3.5 Super VL	–	233	–	–
22	gpt-oss-120b	low	232	–	–
23	Ling 3.0 Flash Fin	–	226	–	–
24	Gemini 3.6 Flash	high	224	–	–
25	Gemini 3.5 Flash	medium	222	–	–
26	Gemini 3 Flash	–	220	–	–
27	Granite 4.2 3B	–	219	–	–
28	GPT-4.1 Nano	none	218	–	–
29	Ling 3.1 Flash	–	215	–	–
30	Ministral 3 3B	none	214	–	–
31	Gemini 2.5 Flash	–	211	–	–
32	NVIDIA Nemotron Nano 9B V2 (Reasoning)	–	209	–	–
33	Qwen 3.7 Max	max	206	–	–
34	NVIDIA Nemotron Nano 12B v2 VL (Non-reasoning)	none	204	–	–
35	Inkling Small	–	200	–	–
36	Step 3.7 Flash	high	199	–	–
37	Inkling	xhigh	192	–	–
38	Nova 2.0 Lite (Medium)	medium	190	–	–
39	Muse Spark 1.1	xhigh	188	–	–
40	o3 mini	high	187	–	–
41	GPT-5 Codex	high	186	–	–
42	Muse Wonder	max	184	–	–
43	Nova 2.0 Lite (High)	high	184	–	–
44	Nova 2.0 Lite (Low)	low	183	–	–
45	Mistral Small 4	–	181	–	–
46	DeepSeek V4.1 Flash	max	180	–	–
47	Nova 2.0 Lite (Non-reasoning)	none	180	–	–
48	Command A	–	178	–	–
49	DeepSeek V4 Flash	max	175	–	–
50	DeepSeek V4 Flash 0731	max	174	–	–
51	NVIDIA Nemotron Nano 9B V2 (Non-reasoning)	none	170	–	–
52	Nova Lite	none	169	–	–
53	GPT-5.1 Codex Mini	high	167	–	–
54	Quasar 438B (Max, Based on GLM-5.2)	max	167	–	–
55	GPT-5.4 Nano	medium	166	–	–
56	Qwen 3.5 35B-A3B	none	166	–	–
57	Mistral Medium 3.5	high	165	–	–
58	Qwen 3	–	163	–	–
59	GPT-4o	none	160	–	–
60	GPT-5.5 Instant	–	159	–	–
61	Solar Pro 3	high	159	–	–
62	GPT-4.1	none	158	–	–
63	Qwen 3 Coder	none	156	–	–
64	Nemotron 3 Ultra	max	156	–	–
65	o4 mini	high	151	–	–
66	Gemini 3.8 Flash	high	147	–	–
67	GPT-4.1 Mini	none	146	–	–
68	Nemotron 3 Super 120B A12B (Reasoning)	–	146	–	–
69	GPT-5 Nano	high	146	–	–
70	Kimi K2	–	145	–	–
71	Qwen 3.6 35B-A3B	none	144	–	–
72	Step 3.5 Flash 2603	–	143	–	–
73	Sonnet 5.5	max	141	–	–
74	DeepSeek R1	–	141	–	–
75	Qwen 3.5 122B-A10B	none	141	–	–
76	GLM 5.2	none	140	–	–
77	Gemini 3.1 Pro	–	140	–	–
78	Llama 3.1 8b	none	139	–	–
79	GPT-5 Mini	high	138	–	–
80	GPT-5.6 Terra	max	137	–	–
81	Nova 2.0 Pro Preview (Low)	low	131	–	–
82	Granite 4.2 8B	–	130	–	–
83	o3	–	130	–	–
84	GPT-5.6 Luna	max	130	–	–
85	Nova Pro	none	130	–	–
86	Nova 2.0 Pro Preview (Medium)	medium	129	–	–
87	Muse Glimmer (High)	high	128	–	–
88	Step 3.5 Flash	–	128	–	–
89	GPT-5.1 Codex	high	127	–	–
90	GPT-5.2 Codex	xhigh	125	–	–
91	GPT-6 Luna	max	124	–	–
92	DeepSeek V3.1	–	121	–	–
93	Gemma 4 12B	none	120	–	–
94	Gemini 2.5 Pro	–	119	–	–
95	Agnes 2.5 Pro Beta	–	118	–	–
96	GPT-5.1	high	118	–	–
97	Ring-2.6-1T	xhigh	117	–	–
98	Qwen 3 30B-A3B	none	115	–	–
99	Nova 2.0 Pro Preview (Non-reasoning)	none	112	–	–
100	Grok 4.20	–	110	–	–
101	GPT-6 Sol	max	108	–	–
102	Ministral 3 8B	none	107	–	–
103	GPT-4o mini	none	107	–	–
104	DeepSeek V3	none	105	–	–
105	GPT-5	high	104	–	–
106	Muse Spark Open (Xhigh)	xhigh	103	–	–
107	Grok 4.3	medium	102	–	–
108	Grok 4.7	xhigh	102	–	–
109	GPT-5.5	xhigh	101	–	–
110	GPT-5.3 Codex	xhigh	101	–	–
111	Grok 3	none	100	–	–
112	GLM 4.7 Flash	none	100	–	–
113	Gemini 4	high	100	–	–
114	North Mini Code	–	99	–	–
115	Qwen 3.5 9B	none	98	–	–
116	Gemma 4 26B	none	98	–	–
117	Haiku 4.5	–	97	–	–
118	Opus 5.5	max	97	–	–
119	MiniMax M3	–	95	–	–
120	GPT-5.6 Sol	max	95	–	–
121	GPT-5.4	xhigh	95	–	–
122	Qwen 2.5 Turbo	none	94	–	–
123	Llama 3.3 70b	none	93	–	–
124	DeepSeek V4 Pro 0424	none	93	–	–
125	Step 5 Preview	–	91	–	–
126	Ministral 3 14B	none	89	–	–
127	DeepSeek V4 Pro 0813	none	89	–	–
128	Hy3	–	89	–	–
129	Grok 4	–	89	–	–
130	GLM 4.6 V	–	88	–	–
131	Reka Flash 3	–	86	–	–
132	Qwen 3.5 27B	none	85	–	–
133	Solar Pro 4	–	82	–	–
134	Llama 4 Scout	none	81	–	–
135	Sonnet 5	max	81	–	–
136	Solar Mini 4	–	81	–	–
137	GPT-5.2	xhigh	80	–	–
138	Mistral Large 3	none	79	–	–
139	Grok Build 0.1	–	77	–	–
140	Granite 4.2 30B	–	76	–	–
141	GLM 4.5 Air	–	76	–	–
142	Magistral Small 1.2	–	74	–	–
143	GLM 5.3	max	74	–	–
144	Mistral Large 4	–	71	–	–
145	Fable 5.1	max	71	–	–
146	Fable 5	max	71	–	–
147	GLM 5	–	70	–	–
148	LongCat 2.0	–	70	–	–
149	GLM 5.1	–	70	–	–
150	GLM 5.3 Flash	max	68	–	–
151	MiniMax M2	–	68	–	–
152	MiniMax M1	–	67	–	–
153	Opus 4.5	–	67	–	–
154	MiniMax M2.5	–	67	–	–
155	Opus 4.8	max	66	–	–
156	Grok 3 Mini	high	64	–	–
157	Sonnet 4.6	max	63	–	–
158	Grok 4.6	high	62	–	–
159	MiniMax M2.1	–	62	–	–
160	GPT-6.1 Sol	max	61	–	–
161	Grok 4.5	high	60	–	–
162	GLM 4.5 V	–	60	–	–
163	Opus 5	max	60	–	–
164	MiniMax M2.7	–	59	–	–
165	Opus 4.7	max	58	–	–
166	Llama 4 Maverick	none	58	–	–
167	Qwen 3.5 397B-A17B	none	56	–	–
168	Qwen 3.6 Plus	–	56	–	–
169	Qwen 3.7 Plus	–	56	–	–
170	Kimi K2.5	none	56	–	–
171	Qwen 3.6 27B	–	56	–	–
172	Mistral Small 3	none	55	–	–
173	GPT-6 Astra	max	55	–	–
174	Qwen 3 32B	none	53	–	–
175	MiMo V2.6 Flash	–	53	–	–
176	Sonnet 4.5	–	51	–	–
177	Qwen 3.8 Flash	–	51	–	–
178	Kimi K2.7 Code	–	50	–	–
179	Gemma 4 31B	none	50	–	–
180	Qwen 3 14B	none	50	–	–
181	MiMo V2.6 Pro	–	48	–	–
182	Kimi K2.6	–	48	–	–
183	Opus 4.6	max	47	–	–
184	DeepSeek V4 Flash 0420	max	47	–	–
185	Gemma 4 E4B (Reasoning)	–	47	–	–
186	Phi-4 Mini Instruct	none	45	–	–
187	Qwen 3 235B-A22B	none	45	–	–
188	Mistral Medium 3	none	45	–	–
189	Granite 3.3 8B (Non-reasoning)	none	44	–	–
190	Kimi K3	max	44	–	–
191	Qwen 3.8 27B	none	43	–	–
192	Gemma 4 E4B (Non-reasoning)	none	42	–	–
193	Phi-4	none	42	–	–
194	Qwen 3.8 Max	max	39	–	–
195	Qwen 3.8 2.4T-A95B	–	39	–	–
196	GLM 4.7	none	38	–	–
197	Seed-OSS-36B-Instruct	–	38	–	–
198	MiMo V2.5 Pro	–	38	–	–
199	GLM 4.6	–	37	–	–
200	Llama 3.1 405b	–	35	–	–
201	MiMo V2.5	–	34	–	–
202	Mistral Small 3.2	none	33	–	–
203	Llama 3.1 70b	none	33	–	–
204	GPT-4	none	32	–	–
205	Mistral Small 3.1	none	30	–	–
206	Qwen 2.5 72B	none	30	–	–
207	Mixtral 8x7B Instruct	none	24	–	–
208	Llama 3.2 11b	none	23	–	–
209	Ling 3.0 Tiny	–	15	–	–
210	Qwen 3.5 4B	–	12	–	–
211	Ling-flash-2.0	none	5	–	–