Dataset Viewer
Auto-converted to Parquet Duplicate
date
stringdate
2026-07-23 00:00:00
2026-08-17 00:00:00
provider
stringclasses
45 values
region
stringclasses
4 values
probe_type
stringclasses
2 values
metric
stringclasses
2 values
p50_ms
float64
11.9
2.8k
p95_ms
float64
44.4
9.06k
samples
int64
8
292
success_pct
float64
69
100
2026-07-23
ai21
ap-tokyo
network
ttfb
304.5
400.6
112
100
2026-07-23
ai21
eu-hetzner
network
ttfb
303.7
501.8
114
100
2026-07-23
ai21
sa-east
network
ttfb
308.9
408.9
112
100
2026-07-23
ai21
us-central
network
ttfb
167.9
276.4
112
100
2026-07-23
aleph-alpha
ap-tokyo
network
ttfb
572.2
638.2
112
100
2026-07-23
aleph-alpha
eu-hetzner
network
ttfb
101.3
293.8
114
100
2026-07-23
aleph-alpha
sa-east
network
ttfb
482.1
564.4
112
100
2026-07-23
aleph-alpha
us-central
network
ttfb
301.9
423.1
112
100
2026-07-23
anthropic
ap-tokyo
network
ttfb
247.2
414.8
135
100
2026-07-23
anthropic
eu-hetzner
network
ttfb
199.1
401.9
149
100
2026-07-23
anthropic
sa-east
network
ttfb
211.8
608.1
135
100
2026-07-23
anthropic
us-central
network
ttfb
107.9
237.3
135
100
2026-07-23
baichuan
ap-tokyo
network
ttfb
178.4
201
112
99.11
2026-07-23
baichuan
eu-hetzner
network
ttfb
178
244.1
114
100
2026-07-23
baichuan
sa-east
network
ttfb
337.8
356.2
112
100
2026-07-23
baichuan
us-central
network
ttfb
233.9
254.6
112
100
2026-07-23
baseten
ap-tokyo
network
ttfb
155.8
168.3
112
100
2026-07-23
baseten
eu-hetzner
network
ttfb
196.8
289
114
100
2026-07-23
baseten
sa-east
network
ttfb
150.2
158
112
100
2026-07-23
baseten
us-central
network
ttfb
54.9
67.2
112
100
2026-07-23
cerebras
ap-tokyo
network
ttfb
201.3
252.8
135
100
2026-07-23
cerebras
eu-hetzner
network
ttfb
198.3
397.3
149
100
2026-07-23
cerebras
sa-east
network
ttfb
186.7
236.6
135
100
2026-07-23
cerebras
us-central
network
ttfb
63.7
130.3
135
100
2026-07-23
cohere
ap-tokyo
network
ttfb
166.9
202.4
135
100
2026-07-23
cohere
eu-hetzner
network
ttfb
199.1
302.9
149
100
2026-07-23
cohere
sa-east
network
ttfb
158
175.5
135
100
2026-07-23
cohere
us-central
network
ttfb
60.5
110.3
135
100
2026-07-23
deepinfra
ap-tokyo
network
ttfb
502.2
684.4
112
100
2026-07-23
deepinfra
eu-hetzner
network
ttfb
597.1
1,092.9
114
100
2026-07-23
deepinfra
sa-east
network
ttfb
589.1
715
112
100
2026-07-23
deepinfra
us-central
network
ttfb
229.4
684.1
112
100
2026-07-23
deepseek
ap-tokyo
network
ttfb
169.7
237.8
135
100
2026-07-23
deepseek
eu-hetzner
network
ttfb
298.8
501.4
149
99.33
2026-07-23
deepseek
sa-east
network
ttfb
425.5
453.7
135
100
2026-07-23
deepseek
us-central
network
ttfb
307.1
371.8
135
100
2026-07-23
doubao
ap-tokyo
network
ttfb
255.4
844.8
112
100
2026-07-23
doubao
eu-hetzner
network
ttfb
354.3
805.1
114
100
2026-07-23
doubao
sa-east
network
ttfb
680.9
1,573.2
112
100
2026-07-23
doubao
us-central
network
ttfb
478.3
1,121.2
112
100
2026-07-23
ernie
ap-tokyo
network
ttfb
137.8
143.7
112
100
2026-07-23
ernie
eu-hetzner
network
ttfb
279.9
284.9
114
100
2026-07-23
ernie
sa-east
network
ttfb
410.1
421.5
112
100
2026-07-23
ernie
us-central
network
ttfb
271
278.2
112
100
2026-07-23
featherless
ap-tokyo
network
ttfb
505.4
2,134.9
112
100
2026-07-23
featherless
eu-hetzner
network
ttfb
700.1
1,102.6
114
100
2026-07-23
featherless
sa-east
network
ttfb
744.9
3,520
112
99.11
2026-07-23
featherless
us-central
network
ttfb
351.5
1,356.5
112
100
2026-07-23
fireworks
ap-tokyo
network
ttfb
11.9
44.4
135
100
2026-07-23
fireworks
eu-hetzner
network
ttfb
98
202
149
100
2026-07-23
fireworks
sa-east
network
ttfb
263.8
305.7
135
100
2026-07-23
fireworks
us-central
network
ttfb
25.6
66.5
135
100
2026-07-23
glm
ap-tokyo
network
ttfb
355.6
387.7
117
100
2026-07-23
glm
eu-hetzner
network
ttfb
801.3
1,494.8
119
100
2026-07-23
glm
sa-east
network
ttfb
648.9
656.5
117
100
2026-07-23
glm
us-central
network
ttfb
386.7
423.8
117
100
2026-07-23
google
ap-tokyo
network
ttfb
58
97.4
135
100
2026-07-23
google
eu-hetzner
network
ttfb
98.3
199
149
100
2026-07-23
google
sa-east
network
ttfb
151.8
303.2
135
100
2026-07-23
google
us-central
network
ttfb
37.6
92.8
135
100
2026-07-23
groq
ap-tokyo
network
ttfb
146.7
187.9
135
100
2026-07-23
groq
eu-hetzner
network
ttfb
296.6
400.7
149
100
2026-07-23
groq
sa-east
network
ttfb
226.8
262.9
135
100
2026-07-23
groq
us-central
network
ttfb
114.8
183.8
135
100
2026-07-23
hunyuan
ap-tokyo
network
ttfb
1,118.1
1,148.8
112
100
2026-07-23
hunyuan
eu-hetzner
network
ttfb
1,539.7
1,577.4
114
100
2026-07-23
hunyuan
sa-east
network
ttfb
1,653.5
1,671.2
112
100
2026-07-23
hunyuan
us-central
network
ttfb
1,378
1,388.2
112
100
2026-07-23
hyperbolic
ap-tokyo
network
ttfb
493.1
550.4
112
100
2026-07-23
hyperbolic
eu-hetzner
network
ttfb
300.9
798
114
100
2026-07-23
hyperbolic
sa-east
network
ttfb
800.4
887
112
100
2026-07-23
hyperbolic
us-central
network
ttfb
229.8
282.2
112
100
2026-07-23
iflytek
ap-tokyo
network
ttfb
224.4
842.7
112
100
2026-07-23
iflytek
eu-hetzner
network
ttfb
299.4
727.7
114
100
2026-07-23
iflytek
sa-east
network
ttfb
638.2
1,533.2
112
100
2026-07-23
iflytek
us-central
network
ttfb
427.1
1,532.5
112
99.11
2026-07-23
inference-net
ap-tokyo
network
ttfb
529
970
112
100
2026-07-23
inference-net
eu-hetzner
network
ttfb
197.1
899.5
114
100
2026-07-23
inference-net
sa-east
network
ttfb
322.4
1,251.3
112
100
2026-07-23
inference-net
us-central
network
ttfb
160.7
205.2
112
100
2026-07-23
kimi
ap-tokyo
network
ttfb
125.7
158.1
117
100
2026-07-23
kimi
eu-hetzner
network
ttfb
299.6
401.1
119
100
2026-07-23
kimi
sa-east
network
ttfb
338.2
374.9
117
100
2026-07-23
kimi
us-central
network
ttfb
264.9
332.8
117
100
2026-07-23
meta-llama
ap-tokyo
network
ttfb
128.2
349.3
112
100
2026-07-23
meta-llama
eu-hetzner
network
ttfb
100.8
290
114
100
2026-07-23
meta-llama
sa-east
network
ttfb
221.8
245
112
100
2026-07-23
meta-llama
us-central
network
ttfb
47.5
110.3
112
100
2026-07-23
mistral
ap-tokyo
network
ttfb
310.5
366.5
135
100
2026-07-23
mistral
eu-hetzner
network
ttfb
99.3
201.6
149
100
2026-07-23
mistral
sa-east
network
ttfb
251.8
290.6
135
100
2026-07-23
mistral
us-central
network
ttfb
182.4
252.1
135
100
2026-07-23
nebius
ap-tokyo
network
ttfb
547
598.4
112
100
2026-07-23
nebius
eu-hetzner
network
ttfb
99.7
296.4
114
100
2026-07-23
nebius
sa-east
network
ttfb
459.5
484.3
112
100
2026-07-23
nebius
us-central
network
ttfb
218
258.4
112
100
2026-07-23
novita
ap-tokyo
network
ttfb
145.4
173.5
112
100
2026-07-23
novita
eu-hetzner
network
ttfb
298.1
496.6
114
100
2026-07-23
novita
sa-east
network
ttfb
236.2
272.7
112
100
2026-07-23
novita
us-central
network
ttfb
67.7
137.9
112
100
End of preview. Expand in Data Studio

LLM Latency Tracker

Independent, continuously measured latency and availability for AI inference API providers, aggregated by day. Covers 45 providers across 4 regions (ap-tokyo, eu-hetzner, sa-east, us-central), built from 1,295,385 raw probes collected between 2026-07-23 and 2026-08-17.

Live rankings and full methodology: llmlatency.dev

How the numbers are produced

Probes run every five minutes from separate network locations and are never routed through a gateway or an aggregator, so the numbers describe the providers themselves rather than a proxy in front of them.

  • Network probe — DNS → TCP → TLS → time to first byte (ttfb).
  • Inference probe — time to first token on a real completion request (ttft).

These are different quantities, an order of magnitude apart, and they are never mixed in one ranking. Percentiles are nearest-rank, identical to the ones shown on the site.

Files

file contents
daily_aggregates.csv one row per date × provider × region × probe type, with p50/p95, sample count and success rate
rankings.json machine-readable snapshot published live at llmlatency.dev

Usage

from datasets import load_dataset

ds = load_dataset("llmlatency/llm-latency-tracker", split="train")
ds.filter(lambda r: r["region"] == "eu-hetzner" and r["probe_type"] == "network")

Or without this dataset at all — there is a client on PyPI:

pip install llmlatency
llmlatency fastest

Limitations, stated up front

Vantage points are cloud data centres, not consumer networks, so absolute values are lower than an end user would see — the comparison between providers is the meaningful part. Provider coverage changes over time as APIs appear and shut down.

Citation

Archived, citable version with a DOI: https://doi.org/10.5281/zenodo.21954788

@dataset{llmlatency_tracker,
  title  = {LLM Latency Tracker: measured latency and uptime for AI inference APIs},
  author = {llmlatency.dev},
  year   = {2026},
  doi    = {10.5281/zenodo.21954788},
  url    = {https://llmlatency.dev}
}

Links

Licence: CC-BY-4.0. Snapshot generated 2026-08-17.

Downloads last month
48