Leaderboard

LMSpeed Latency Leaderboard

Name: Monthly LLM Model Latency Leaderboard | LMSpeed
Creator: LMSpeed
License: https://creativecommons.org/licenses/by/4.0/

Multi-dimensional rankings based on model speed tests and provider health checks. Compare providers, endpoints, and reliability at a glance.

Ranked by median time to first token (resistant to outliers). Lower is better for responsiveness.

Throughput First Token Latency Health

Fairness Notice

To ensure fairness, the system uses AI-powered detection. Suspicious data or cheating will be flagged and removed from the leaderboard.

Top performerRankedHigh confidenceMedium confidenceLow confidence

Last 7 Days Last 30 Days

Archive Month

Apr 2026 Mar 2026 Feb 2026 Jan 2026 Dec 2025 Nov 2025 Oct 2025 Sep 2025 Aug 2025 Jul 2025 Jun 2025 May 2025 Apr 2025 Mar 2025 Feb 2025 Jan 2025 Dec 2024 Nov 2024

Rank	Provider	Model	First Token Latency	Avg tokens per second	Total Tests
1NEW	NVIDIA NIMintegrate.api.nvidia.com	meta/llama-4-maverick-17b-128e-instruct	0.19 s Best: 0.17Worst: 0.31	100.29t/s	15
2NEW	SiliconFlowapi.siliconflow.cn	tencent/Hunyuan-MT-7B	0.24 s Best: 0.21Worst: 0.31	67.44t/s	10
31	NVIDIA NIMintegrate.api.nvidia.com	minimaxai/minimax-m2.1	0.25 s Best: 0.18Worst: 4.31	95.95t/s	10
41	钠 APInaapi.cc	llama3.1-8B	0.41 s Best: 0.37Worst: 0.76	1070.27t/s	15
5NEW	SiliconFlowapi.siliconflow.cn	THUDM/GLM-4-9B-0414	0.46 s Best: 0.29Worst: 4.96	63.75t/s	10
6NEW	I IPv4 Beta LM Studioipv4-beta.kxcym.top:11434	qwen3.5-0.8b	0.46 s Best: 0.45Worst: 1.59	356.07t/s	10
7NEW	NVIDIA NIMintegrate.api.nvidia.com	institute-of-science-tokyo/llama-3.1-swallow-70b-instruct-v0.1	0.47 s Best: 0.44Worst: 0.61	19.05t/s	10
8NEW	NVIDIA NIMintegrate.api.nvidia.com	abacusai/dracarys-llama-3.1-70b-instruct	0.47 s Best: 0.45Worst: 0.60	19.09t/s	10
9NEW	I IPv4 Beta LM Studioipv4-beta.kxcym.top:11434	huihui-ai/Huihui-Qwen3.5-0.8B-abliterated	0.48 s Best: 0.43Worst: 13.12	195.97t/s	10
107	A AI Toolsplatform.aitools.cfd	zhipu/glm-4v-flash	0.49 s Best: 0.34Worst: 0.83	51.56t/s	45
11NEW	SiliconFlowapi.siliconflow.cn	deepseek-ai/DeepSeek-R1-Distill-Qwen-7B	0.49 s Best: 0.39Worst: 6.13	83.11t/s	10
12NEW	NVIDIA NIMintegrate.api.nvidia.com	meta/llama-3.1-405b-instruct	0.49 s Best: 0.23Worst: 32.15	22.72t/s	20
13NEW	K Koru APInew.api.koru.ink	llama3.1-8B	0.50 s Best: 0.47Worst: 7.53	784.80t/s	15
14NEW	NVIDIA NIMintegrate.api.nvidia.com	openai/gpt-oss-120b	0.51 s Best: 0.33Worst: 0.79	201.82t/s	25
15NEW	SiliconFlowapi.siliconflow.cn	Qwen/Qwen3-8B	0.53 s Best: 0.42Worst: 0.70	23.35t/s	10
16NEW	NVIDIA NIMintegrate.api.nvidia.com	minimaxai/minimax-m2.5	0.55 s Best: 0.23Worst: 15.46	64.27t/s	30
17	A AI Toolsplatform.aitools.cfd	qwen/qwen2.5-7b	0.56 s Best: 0.34Worst: 2.01	93.01t/s	140
18NEW	NVIDIA NIMintegrate.api.nvidia.com	moonshotai/kimi-k2-instruct	0.57 s Best: 0.43Worst: 0.75	41.18t/s	10
19NEW	Supabase AI Proxyttknrllwjndwdtycqqfv.supabase.co	llama3.1-8b	0.58 s Best: 0.53Worst: 0.64	2365.78t/s	10
20NEW	A AI Toolsplatform.aitools.cfd	qwen/qwen2.5-7b	0.65 s Best: 0.37Worst: 2.22	96.76t/s	10
213	A AI Toolsplatform.aitools.cfd	zhipu/glm-4-flash	0.67 s Best: 0.51Worst: 1.66	29.19t/s	2525
22NEW	NVIDIA NIMintegrate.api.nvidia.com	qwen/qwen3-coder-480b-a35b-instruct	0.72 s Best: 0.62Worst: 2.56	51.33t/s	25
23NEW	温温云sxtuyxrxcgim.ap-northeast-1.clawcloudrun.com	moonshotai/kimi-k2-instruct-0905	0.74 s Best: 0.70Worst: 1.10	65.70t/s	10
24NEW	Q QYES AIai.qyes.top	GLM-4-Flash-250414	0.74 s Best: 0.58Worst: 1.39	38.40t/s	10
25NEW	G GankInterview LLMllm.gankinterview.com	gemini-2.5-flash-lite	0.75 s Best: 0.61Worst: 1.05	251.28t/s	10
2625	NVIDIA NIMintegrate.api.nvidia.com	qwen/qwen3.5-122b-a10b	0.77 s Best: 0.26Worst: 14.82	23.24t/s	55
27NEW	X XShuLab Sub2APIapi.xshulab.com	gpt-5.4-mini	0.80 s Best: 0.63Worst: 1.54	162.55t/s	10
2815	星见雅 APIapi.xinjianya.top	grok-4.20-beta	0.86 s Best: 0.53Worst: 1.99	46.28t/s	15
29NEW	Vercel AI Gatewayai-gateway.vercel.sh	google/gemini-2.5-flash-lite	0.89 s Best: 0.64Worst: 1.99	209.12t/s	10
30NEW	ModelScopems-ens-1f4a9445-d0e7.api-inference.modelscope.cn	kgiser/gpu_gpt_5	0.94 s Best: 0.74Worst: 2.18	30.27t/s	10
31NEW	素素墨APIapifree.rensumo.top	openai/gpt-oss-20b	1.01 s Best: 0.77Worst: 1.41	272.22t/s	10
32	SiliconFlowapi.siliconflow.cn	deepseek-ai/DeepSeek-V3.2	1.04 s Best: 0.68Worst: 5.26	22.93t/s	15
33	阿里云百炼 DashScopecoding.dashscope.aliyuncs.com	kimi-k2.5	1.08 s Best: 0.89Worst: 3.62	42.23t/s	65
348	Krioraapi.kriora.com	kimi-k2.5	1.09 s Best: 0.92Worst: 1.78	1073.96t/s	15
35NEW	M Mars HKmars-hk.duckdns.org:38317	GLM-4.7	1.13 s Best: 0.88Worst: 1.52	129.35t/s	15
365	DeepSeekapi.deepseek.com	deepseek-chat	1.16 s Best: 0.92Worst: 1.47	32.12t/s	15
37NEW	I IPv4 Beta LM Studioipv4-beta.kxcym.top:11434	qwen3.5-instruct	1.17 s Best: 0.98Worst: 2.25	56.39t/s	10
38NEW	Sub2APIapi.243706.xyz	gpt-5.4	1.23 s Best: 0.88Worst: 2.98	49.20t/s	25
39NEW	心流apis.iflow.cn	qwen3-coder-plus	1.26 s Best: 0.81Worst: 2.10	38.40t/s	10
4017	A AI Toolsplatform.aitools.cfd	qwen/qwen3-8b	1.27 s Best: 0.57Worst: 37.10	25.26t/s	85
41NEW	QuicklyAPIsub.jlypx.de	gpt-5.4(high)	1.28 s Best: 1.00Worst: 6.15	51.89t/s	15
42NEW	晴辰云gpt.qt.cool	gpt-5.1-codex-mini	1.32 s Best: 1.05Worst: 2.51	186.77t/s	10
43NEW	OpenRouteropenrouter.ai	arcee-ai/trinity-large-preview:free	1.32 s Best: 0.84Worst: 1.83	6.42t/s	15
44NEW	SiliconFlowapi.siliconflow.cn	deepseek-ai/DeepSeek-V3	1.34 s Best: 0.97Worst: 9.97	18.82t/s	15
45NEW	火山引擎 Arkark.cn-beijing.volces.com	deepseek-v3-2-251201	1.46 s Best: 0.98Worst: 4.56	29.79t/s	10
46NEW	星见雅 APIapi.xinjianya.top	李/gpt-5.3-codex	1.50 s Best: 1.37Worst: 1.70	53.69t/s	10
47NEW	晴辰云gpt.qt.cool	gpt-5.2-codex	1.58 s Best: 1.27Worst: 5.75	56.92t/s	11
48NEW	星见雅 APIapi.xinjianya.top	gpt-5.2	1.58 s Best: 1.08Worst: 6.67	53.95t/s	10
49NEW	S Sliamapi.sliam.site	gemini-3.1-flash-lite-preview	1.58 s Best: 1.17Worst: 1.99	182.91t/s	10
50NEW	MiniMaxapi.minimaxi.com	MiniMax-M2.7-highspeed	1.59 s Best: 0.67Worst: 3.49	47.25t/s	20

Throughputt/s

HealthAvg probe latency

Model

Provider