Speed: output tokens per second

Median output throughput measured across healthy providers over the last 30 minutes of live traffic (OpenRouter).

557 models · data as of 9 Sept 2026

1. 1Celeris-11360
2. 2Mercury 2744
3. 3Gemini 2.5 Flash-Lite378
4. 4Gemini 3.5 Flash Lite369
5. 5Gemini 3.5 Flash Lite369
6. 6Gemini 3.5 Flash Lite369
7. 7Gemini 3.5 Flash Lite369
8. 8Gemini 3.5 Flash Lite369
9. 9HyperNova 60B 2605 (high, based on gpt-oss-120b)345
10. 10Grok 4.20 Multi Agent329
11. 11Trinity Large321
12. 12Gemini 3.7 Flash303
13. 13Gemini 3.1 Flash Lite303
14. 14Gemini 3.1 Flash Lite303
15. 15Gemini 3.1 Flash Lite303

SearchAll providersAlibabaAmazonAnt GroupAnt-groupAnthropicArcee AIByteDanceCelerisCohereDeepSeekGoogleIBMInceptionInclusionAILiquid AILongCatMetaMicrosoftMiniMaxMistral AIMoonshot AIMultiverse ComputingNex AGINous ResearchNVIDIAOpenAIPerplexitySapiens AIStepFunThinking MachinesUpstagexAIXiaomiZhipu AIOpen weights557 of 557

#	Model	tok/s ↓	First token ↕
1	CCeleris-1Celeris	1360	0.6 s
2	Mercury 2Inception	744	0.9 s
3	Gemini 2.5 Flash-LiteGoogle	378	0.3 s
4	Gemini 3.5 Flash LiteGoogle	369	1.1 s
5	Gemini 3.5 Flash LiteGoogle	369	1.1 s
6	Gemini 3.5 Flash LiteGoogle	369	1.1 s
7	Gemini 3.5 Flash LiteGoogle	369	1.1 s
8	Gemini 3.5 Flash LiteGoogle	369	1.1 s
9	MHyperNova 60B 2605 (high, based on gpt-oss-120b)Multiverse Computingopen	345	–
10	Grok 4.20 Multi AgentxAI	329	6.7 s
11	ATrinity LargeArcee AI	321	0.2 s
12	Gemini 3.7 FlashGoogle	303	1.5 s
13	Gemini 3.1 Flash LiteGoogle	303	0.6 s
14	Gemini 3.1 Flash LiteGoogle	303	0.6 s
15	Gemini 3.1 Flash LiteGoogle	303	0.6 s
16	Gemini 3.1 Flash LiteGoogle	303	0.6 s
17	Gemini 3.7 FlashGoogle	297	1.5 s
18	Gemini 3.7 FlashGoogle	297	1.5 s
19	ILing 3.0 FlashInclusionAIopen	290	0.7 s
20	Gemini 2.5 Flash-LiteGoogle	290	0.7 s
21	Gemini 3.7 FlashGoogle	282	1.5 s
22	Nemotron 3.5 LightningNVIDIAopen	277	1.1 s
23	Gemini 3.8 FlashGoogle	273	1.6 s
24	Gemini 3.8 FlashGoogle	273	1.6 s
25	Gemini 3.8 FlashGoogle	273	1.6 s
26	Gemini 3.8 FlashGoogle	273	1.6 s
27	Nova MicroAmazon	267	0.8 s
28	gpt-oss-20bOpenAIopen	249	0.4 s
29	Command A+Cohereopen	235	–
30	Qwen3.5 Omni FlashAlibaba	235	1.8 s
31	nemotron-3-nano-30b-a3bNVIDIAopen	231	1.0 s
32	Muse Spark 1.3Meta	226	3.7 s
33	Muse Spark 1.3Meta	226	3.7 s
34	gpt-oss-120bOpenAIopen	224	0.5 s
35	Muse Spark 1.2Meta	219	3.9 s
36	Muse Spark 1.2Meta	219	3.9 s
37	GPT-5.4 miniOpenAI	218	1.3 s
38	GPT-5.4 miniOpenAI	218	1.3 s
39	GPT-5.4 miniOpenAI	218	1.3 s
40	GPT-5.4 miniOpenAI	218	1.3 s
41	Step 3.5 FlashStepFunopen	216	1.3 s
42	Granite 4.2 3BIBMopen	215	–
43	o3-miniOpenAI	213	5.5 s
44	o3-miniOpenAI	213	5.5 s
45	o3-miniOpenAI	213	5.5 s
46	o3-miniOpenAI	213	5.5 s
47	Gemini 2.5 FlashGoogle	211	0.4 s
48	Gemini 3.5 FlashGoogle	209	1.7 s
49	Gemini 3.5 FlashGoogle	209	1.7 s
50	Gemini 3.5 FlashGoogle	209	1.7 s
51	Gemini 3.5 FlashGoogle	209	1.7 s
52	Gemini 3 FlashGoogle	202	0.8 s
53	Gemini 3 FlashGoogle	202	1.5 s
54	Gemini 3 FlashGoogle	202	0.8 s
55	Gemini 3 FlashGoogle	202	0.8 s
56	Gemini 3 FlashGoogle	202	0.8 s
57	gpt-oss-120bOpenAIopen	202	0.5 s
58	gpt-oss-120bOpenAIopen	202	0.5 s
59	gpt-oss-120bOpenAIopen	202	0.5 s
60	Gemini 3 FlashGoogle	200	0.8 s

Show 100 more