Provider throughput

Median provider output throughput in tokens per second across OpenRouter endpoints (3d).

178models

–Open Index weight

–last updated

OpenRouterpublisher

#	Model	Effort	Score	Cost / task	Harness
1	Muse Spark 1.1	max	198	–	–
2	Llama 3.2 3b	none	173	–	–
3	Gemini 2.5 Flash Lite	none	160	–	–
4	Haiku 5.5	max	142	–	–
5	Gemini 3.5 Flash	xhigh	131	–	–
6	Gemini 3.5 Flash-Lite	xhigh	128	–	–
7	Gemini 3.6 Flash	xhigh	127	–	–
8	o3 mini	high	125	–	–
9	Inkling Small	xhigh	121	–	–
10	Gemini 3.1 Flash-Lite	high	121	–	–
11	gpt-oss-120b	high	120	–	–
12	Codestral 2508	–	117	–	–
13	Mistral Small 4	high	115	–	–
14	o4 mini	high	111	–	–
15	Qwen 2.5 7B	–	110	–	–
16	Muse Spark 1.2	max	109	–	–
17	Mistral Medium 3.1	none	108	–	–
18	Qwen 3.6 Flash	max	107	–	–
19	Llama 3.2 1b	none	107	–	–
20	Mistral Medium 3.5	high	105	–	–
21	Llama 3.1 8b	medium	104	–	–
22	Gemini 3.1 Flash	–	103	–	–
23	GPT-5.6 Luna Pro	–	102	–	–
24	Mistral Medium 3	none	102	–	–
25	GPT-5.6 Sol Pro	max	102	–	–
26	GPT-5.4 Mini	xhigh	102	–	–
27	gpt-oss-20b	high	101	–	–
28	Sonnet 5.5	max	99	–	–
29	DeepSeek V4.1 Flash	max	95	–	–
30	Gemini 3 Flash	high	95	–	–
31	GPT-5.4 Nano	xhigh	92	–	–
32	Qwen 3 30B-A3B	none	92	–	–
33	GPT-5 Mini	high	91	–	–
34	Gemini 3.8 Flash	max	90	–	–
35	DeepSeek V4 Flash	max	90	–	–
36	Gemma 3 4B	none	89	–	–
37	Gemini 3.1 Pro	high	89	–	–
38	Grok Build 0.1	–	87	–	–
39	GPT-4.1 Nano	high	87	–	–
40	Gemini 2.5 Pro	high	85	–	–
41	Gemini 2.5 Flash	medium	85	–	–
42	Qwen 3.5 Flash	max	84	–	–
43	GPT-5.6 Terra Pro	–	83	–	–
44	Gemini 3.7 Flash	max	82	–	–
45	GPT-5 Nano	high	82	–	–
46	GPT-5.6 Luna	max	82	–	–
47	GPT-4.1	high	81	–	–
48	Inkling	max	80	–	–
49	o1	high	80	–	–
50	Grok 4.3	xhigh	80	–	–
51	Muse Spark 1.3	max	78	–	–
52	Grok 4.20	xhigh	78	–	–
53	Devstral 2	none	78	–	–
54	GPT-3.5	high	78	–	–
55	Haiku 4.5	xhigh	77	–	–
56	GLM 5.2	max	77	–	–
57	GPT-6 Luna Pro	max	77	–	–
58	Opus 5.5	max	77	–	–
59	GLM 5.3	max	76	–	–
60	DeepSeek V3.1	none	76	–	–
61	Opus 4.8	max	75	–	–
62	Muse Wonder	max	75	–	–
63	Sonnet 5	max	75	–	–
64	Gemini 3 Pro	high	73	–	–
65	GPT-5.6 Terra	max	72	–	–
66	GLM 5.3 Flash	max	71	–	–
67	GPT-4o	high	71	–	–
68	MiniMax M3	max	69	–	–
69	Qwen 3.5 122B-A10B	none	69	–	–
70	Qwen 3.7 Max	max	68	–	–
71	GPT-4.1 Mini	high	68	–	–
72	GPT-5.6 Sol	max	68	–	–
73	Opus 5	max	67	–	–
74	Grok 4.7	max	66	–	–
75	GPT-4o mini	high	66	–	–
76	Qwen 3.8	xhigh	65	–	–
77	Qwen 3.5 35B-A3B	low	64	–	–
78	Mistral Large 4	high	63	–	–
79	Mistral Large	none	63	–	–
80	GPT-6 Sol Pro	max	62	–	–
81	DeepSeek V4 Flash 0731	max	62	–	–
82	Fable 5.1	max	61	–	–
83	GPT-5.1	high	61	–	–
84	GPT-5.5	xhigh	58	–	–
85	Mistral Large 3	none	58	–	–
86	GPT-6 Sol	max	57	–	–
87	Qwen 3.7 Flash	max	57	–	–
88	Qwen 3.6 35B-A3B	max	57	–	–
89	GLM 5	high	56	–	–
90	GPT-5.4	xhigh	55	–	–
91	GPT-6 Luna	max	55	–	–
92	Qwen 3 Coder	medium	55	–	–
93	GPT-5	high	55	–	–
94	GPT-4	high	55	–	–
95	Opus 4.7	max	55	–	–
96	GPT-5.1 Codex Mini	high	54	–	–
97	o3	high	54	–	–
98	Qwen 3.5 Plus	none	54	–	–
99	GPT-5.3 Codex	xhigh	53	–	–
100	GPT-5.2	xhigh	53	–	–
101	Mistral Large 2407	–	53	–	–
102	Qwen 3.5 9B	low	52	–	–
103	Grok 4.5	max	52	–	–
104	Sonnet 4.6	max	52	–	–
105	Qwen 3 14B	none	52	–	–
106	GLM 4.5 V	none	51	–	–
107	GPT-5.1 Codex	max	50	–	–
108	Grok 4.6	max	50	–	–
109	Kimi K3	max	50	–	–
110	Qwen 3.7 Plus	high	50	–	–
111	Qwen 3.8 Flash	max	49	–	–
112	GLM 4.5 Air	–	49	–	–
113	Mistral Small 3	none	49	–	–
114	Fable 5	max	49	–	–
115	DeepSeek V4 Flash 0423	–	49	–	–
116	MiniMax M2.5	high	49	–	–
117	Qwen 3.5 397B-A17B	max	49	–	–
118	Opus 4.5	high	49	–	–
119	Qwen 3.8 27B	max	48	–	–
120	Hy3	xhigh	48	–	–
121	Gemma 4 26B	xhigh	48	–	–
122	Llama 3.1 70b	none	48	–	–
123	GLM 4.7 Flash	none	47	–	–
124	Llama 3.3 70b	none	47	–	–
125	DeepSeek V4 Pro 0423	–	46	–	–
126	MiniMax M2.1	–	46	–	–
127	MiniMax M2.7	high	45	–	–
128	GPT-6.1 Sol Pro	–	44	–	–
129	DeepSeek V4 Pro 0813	max	44	–	–
130	Qwen 3.6 27B	max	44	–	–
131	GLM 5.1	max	44	–	–
132	MiniMax M2	–	44	–	–
133	Gemini 4	max	44	–	–
134	GPT-6 Astra Pro	max	43	–	–
135	Qwen 3.6 Plus	max	43	–	–
136	GPT-5.2 Codex	xhigh	43	–	–
137	Qwen 3	high	43	–	–
138	Kimi K2.5	max	42	–	–
139	Hy4	–	41	–	–
140	Kimi K2.7 Code	max	41	–	–
141	Mistral	none	41	–	–
142	MiniMax M1	–	41	–	–
143	GPT-6 Astra	max	40	–	–
144	Mistral Small 3.2	none	40	–	–
145	Opus 4.6	max	39	–	–
146	Sonnet 4.5	high	38	–	–
147	Kimi K2	medium	37	–	–
148	GLM 5 V	high	37	–	–
149	Qwen 3.8 2.4T-A95B	–	36	–	–
150	Qwen 3 235B-A22B	xhigh	36	–	–
151	Mistral Small 3.1	none	36	–	–
152	Qwen 2.5	–	36	–	–
153	Qwen 3.8 Max	max	35	–	–
154	Qwen 3.5 27B	high	35	–	–
155	Kimi K2.6	max	34	–	–
156	GLM 4.6 V	low	34	–	–
157	Llama 4 Maverick	medium	33	–	–
158	GPT-6.1 Sol	max	33	–	–
159	MiMo V2.6 Pro	high	33	–	–
160	Gemma 3 27B	none	31	–	–
161	Qwen 2.5 Coder	none	31	–	–
162	Qwen 3 32B	none	31	–	–
163	Qwen 2.5 72B	none	29	–	–
164	DeepSeek V3.2	high	28	–	–
165	MiMo V2.5 Pro	xhigh	26	–	–
166	Gemma 3 12B	none	26	–	–
167	MiMo V2.6 Flash	high	25	–	–
168	Gemma 4 31B	max	25	–	–
169	GLM 4.6	none	25	–	–
170	GLM 4.5	xhigh	25	–	–
171	MiMo V2.5	xhigh	22	–	–
172	Llama 4 Scout	medium	21	–	–
173	DeepSeek V3	none	20	–	–
174	GPT-5.5 Pro	xhigh	16	–	–
175	Gemma 2 27B	–	14	–	–
176	GLM 4.7	none	13	–	–
177	Opus 4.1	high	11	–	–
178	o3 pro	high	8	–	–