HomeOverview
Data API

AI API latency

Compare endpoint response times and availability across regions.

Download JSON
Overview LatencyUptime IncidentsLast probe · 22 Oct, 08:41 UTC
Fastest typical response
19ms

Fireworks AI · Asia (Tokyo)

Endpoint availability
99.8%

52,992 network probes · last 24h

Providers
46

API endpoints monitored

Regions
4

Independent probe locations

Response time over 24 hours

Typical response (p50 TTFB) · Europe (Germany)

Lower is faster
Hourly p50 edge latency; missing observations are gapsResponse time · ms08817526235009:0015:0021:0003:0008:00OpenAI · 2026-10-21T09:00 UTC · 295 ms p50OpenAI · 2026-10-21T10:00 UTC · 292 ms p50OpenAI · 2026-10-21T11:00 UTC · 300 ms p50OpenAI · 2026-10-21T12:00 UTC · 303 ms p50OpenAI · 2026-10-21T13:00 UTC · 299 ms p50OpenAI · 2026-10-21T14:00 UTC · 296 ms p50OpenAI · 2026-10-21T15:00 UTC · 299 ms p50OpenAI · 2026-10-21T16:00 UTC · 295 ms p50OpenAI · 2026-10-21T17:00 UTC · 301 ms p50OpenAI · 2026-10-21T18:00 UTC · 299 ms p50OpenAI · 2026-10-21T19:00 UTC · 295 ms p50OpenAI · 2026-10-21T20:00 UTC · 289 ms p50OpenAI · 2026-10-21T21:00 UTC · 209 ms p50OpenAI · 2026-10-21T22:00 UTC · 299 ms p50OpenAI · 2026-10-21T23:00 UTC · 267 ms p50OpenAI · 2026-10-22T00:00 UTC · 200 ms p50OpenAI · 2026-10-22T01:00 UTC · 205 ms p50OpenAI · 2026-10-22T02:00 UTC · 297 ms p50OpenAI · 2026-10-22T03:00 UTC · 295 ms p50OpenAI · 2026-10-22T04:00 UTC · 305 ms p50OpenAI · 2026-10-22T05:00 UTC · 225 ms p50OpenAI · 2026-10-22T06:00 UTC · 298 ms p50OpenAI · 2026-10-22T07:00 UTC · 293 ms p50OpenAI · 2026-10-22T08:00 UTC · 293 ms p50Anthropic · 2026-10-21T09:00 UTC · 207 ms p50Anthropic · 2026-10-21T10:00 UTC · 201 ms p50Anthropic · 2026-10-21T11:00 UTC · 201 ms p50Anthropic · 2026-10-21T12:00 UTC · 199 ms p50Anthropic · 2026-10-21T13:00 UTC · 198 ms p50Anthropic · 2026-10-21T14:00 UTC · 200 ms p50Anthropic · 2026-10-21T15:00 UTC · 200 ms p50Anthropic · 2026-10-21T16:00 UTC · 201 ms p50Anthropic · 2026-10-21T17:00 UTC · 199 ms p50Anthropic · 2026-10-21T18:00 UTC · 199 ms p50Anthropic · 2026-10-21T19:00 UTC · 293 ms p50Anthropic · 2026-10-21T20:00 UTC · 204 ms p50Anthropic · 2026-10-21T21:00 UTC · 200 ms p50Anthropic · 2026-10-21T22:00 UTC · 199 ms p50Anthropic · 2026-10-21T23:00 UTC · 199 ms p50Anthropic · 2026-10-22T00:00 UTC · 200 ms p50Anthropic · 2026-10-22T01:00 UTC · 198 ms p50Anthropic · 2026-10-22T02:00 UTC · 290 ms p50Anthropic · 2026-10-22T03:00 UTC · 202 ms p50Anthropic · 2026-10-22T04:00 UTC · 199 ms p50Anthropic · 2026-10-22T05:00 UTC · 198 ms p50Anthropic · 2026-10-22T06:00 UTC · 203 ms p50Anthropic · 2026-10-22T07:00 UTC · 203 ms p50Anthropic · 2026-10-22T08:00 UTC · 198 ms p50Google · 2026-10-21T09:00 UTC · 103 ms p50Google · 2026-10-21T10:00 UTC · 92 ms p50Google · 2026-10-21T11:00 UTC · 104 ms p50Google · 2026-10-21T12:00 UTC · 99 ms p50Google · 2026-10-21T13:00 UTC · 100 ms p50Google · 2026-10-21T14:00 UTC · 100 ms p50Google · 2026-10-21T15:00 UTC · 99 ms p50Google · 2026-10-21T16:00 UTC · 102 ms p50Google · 2026-10-21T17:00 UTC · 100 ms p50Google · 2026-10-21T18:00 UTC · 97 ms p50Google · 2026-10-21T19:00 UTC · 97 ms p50Google · 2026-10-21T20:00 UTC · 99 ms p50Google · 2026-10-21T21:00 UTC · 102 ms p50Google · 2026-10-21T22:00 UTC · 100 ms p50Google · 2026-10-21T23:00 UTC · 98 ms p50Google · 2026-10-22T00:00 UTC · 98 ms p50Google · 2026-10-22T01:00 UTC · 100 ms p50Google · 2026-10-22T02:00 UTC · 101 ms p50Google · 2026-10-22T03:00 UTC · 98 ms p50Google · 2026-10-22T04:00 UTC · 96 ms p50Google · 2026-10-22T05:00 UTC · 98 ms p50Google · 2026-10-22T06:00 UTC · 99 ms p50Google · 2026-10-22T07:00 UTC · 101 ms p50Google · 2026-10-22T08:00 UTC · 96 ms p50
OpenAIAnthropicGoogleUTC · lower is faster
View hourly values

Hourly median (p50), in milliseconds. Times are UTC. A dash means no measurement.

Hourly endpoint response times
Date · UTCOpenAIAnthropicGoogle
10-22 08:0029319896
10-22 07:00293203101
10-22 06:0029820399
10-22 05:0022519898
10-22 04:0030519996
10-22 03:0029520298
10-22 02:00297290101
10-22 01:00205198100
10-22 00:0020020098
10-21 23:0026719998
10-21 22:00299199100
10-21 21:00209200102
10-21 20:0028920499
10-21 19:0029529397
10-21 18:0029919997
10-21 17:00301199100
10-21 16:00295201102
10-21 15:0029920099
10-21 14:00296200100
10-21 13:00299198100
10-21 12:0030319999
10-21 11:00300201104
10-21 10:0029220192
10-21 09:00295207103
Looking for model response time? View time to first token (TTFT)

Speed index

Relative endpoint speed · 4 measured regions

Higher is faster
01Fireworks AI81/10002OpenRouter68/10003Google54/10004meta-llama49/10005baseten40/100

100 matches the fastest endpoint in every measured region. 50 means twice the latency in a region; scores are averaged across regions.

Full rankings & scoring method

Provider rankings

Choose the region closest to your application’s server · last 24 hours

Endpoint latency
How to read these metrics

Typical response · p50Half of successful timed probes were this fast or faster. Lower is better.

Slower responses · p9595% of successful timed probes were this fast or faster. This shows how latency behaves on slower requests.

AvailabilityThe share of successful network probes. Higher is better. Samples show how many probes were made.

TTFB measures the network and API endpoint, before a model generates an answer. It does not measure answer quality or generation speed. Compare model TTFT or read the measurement method. A dash means no timing data.

Fastest typical response first

Asia (Tokyo)

RankProviderTypical responsep50 TTFB · msSlower responsesp95 TTFB · msAvailabilitySuccessful probesSamplesLast 24 hours
01Fireworks AI19 ms60 ms100%288
02SambaNova27 ms73 ms100%288
03upstage59 ms77 ms100%288
04OpenRouter61 ms136 ms100%288
05Google66 ms134 ms100%288
06glm91 ms155 ms100%288
07meta-llama117 ms288 ms100%288
08kimi131 ms221 ms100%288
09ernie135 ms163 ms100%288
10baichuan140 ms518 ms99.0%288
11TypeSafe146 ms166 ms100%288
12novita150 ms214 ms100%288
13qwen159 ms219 ms100%288
14Groq165 ms251 ms100%288
15baseten168 ms194 ms100%288
16xAI170 ms264 ms100%288
17Cohere181 ms223 ms100%288
18DeepSeek182 ms249 ms100%288
19Together AI182 ms351 ms100%288
20yi-01ai191 ms1,282 ms85.8%288
21replicate200 ms480 ms100%288
22Cerebras205 ms285 ms99.7%288
23inference-net206 ms457 ms100%288
24OpenAI209 ms306 ms100%288
25Perplexity219 ms300 ms100%288
26minimax222 ms255 ms100%288
27iflytek234 ms839 ms99.0%288
28Anthropic240 ms322 ms100%288
29sarvam251 ms265 ms100%288
30reka273 ms626 ms100%288
31writer274 ms378 ms100%288
32friendli277 ms508 ms100%288
33targon316 ms334 ms100%288
34Mistral322 ms391 ms100%288
35stepfun345 ms960 ms100%288
36ai21349 ms471 ms100%288
37siliconflow371 ms396 ms100%288
38sensenova441 ms1,128 ms100%288
39deepinfra447 ms977 ms100%288
40nebius479 ms523 ms100%288
41hyperbolic484 ms561 ms100%288
42doubao494 ms1,217 ms99.3%288
43featherless516 ms2,857 ms99.7%288
44nscale541 ms548 ms100%288
45aleph-alpha548 ms583 ms100%288
46hunyuan1,115 ms1,125 ms100%288
46 providersRegion report

Europe (Germany)

RankProviderTypical responsep50 TTFB · msSlower responsesp95 TTFB · msAvailabilitySuccessful probesSamplesLast 24 hours
01Fireworks AI98 ms203 ms99.7%288
02nscale98 ms204 ms99.7%288
03Google99 ms202 ms100%288
04OpenRouter99 ms203 ms100%288
05aleph-alpha100 ms204 ms100%288
06meta-llama100 ms206 ms100%288
07nebius100 ms299 ms100%288
08Mistral102 ms296 ms99.3%288
09inference-net104 ms594 ms99.3%288
10baichuan196 ms497 ms100%288
11TypeSafe198 ms299 ms99.7%288
12Cerebras198 ms305 ms99.3%288
13Cohere199 ms306 ms100%288
14Perplexity199 ms308 ms100%288
15glm199 ms305 ms100%288
16baseten199 ms303 ms100%288
17Anthropic200 ms308 ms98.6%288
18replicate200 ms405 ms99.0%288
19reka206 ms504 ms100%288
20yi-01ai228 ms400 ms100%288
21ernie284 ms303 ms100%288
22OpenAI294 ms490 ms99.3%288
23xAI295 ms400 ms99.3%288
24siliconflow295 ms399 ms100%288
25sarvam295 ms306 ms100%288
26Together AI297 ms496 ms99.3%288
27Groq297 ms405 ms99.3%288
28writer297 ms498 ms100%288
29targon297 ms400 ms100%288
30novita298 ms494 ms99.0%288
31DeepSeek299 ms404 ms100%288
32kimi299 ms497 ms99.0%288
33iflytek300 ms755 ms100%288
34doubao360 ms798 ms100%288
35sensenova394 ms951 ms100%288
36SambaNova396 ms503 ms100%288
37friendli396 ms800 ms100%288
38minimax400 ms600 ms100%288
39ai21400 ms601 ms99.7%288
40qwen400 ms700 ms100%288
41stepfun413 ms1,043 ms100%288
42hyperbolic494 ms899 ms98.6%288
43upstage548 ms602 ms100%288
44deepinfra594 ms1,098 ms100%288
45featherless701 ms2,395 ms99.7%288
46hunyuan1,533 ms1,598 ms100%288
46 providersRegion report

South America (São Paulo)

RankProviderTypical responsep50 TTFB · msSlower responsesp95 TTFB · msAvailabilitySuccessful probesSamplesLast 24 hours
01OpenRouter62 ms101 ms100%288
02Google140 ms240 ms100%288
03baseten153 ms167 ms100%288
04Cohere166 ms200 ms100%288
05TypeSafe174 ms186 ms100%288
06Cerebras188 ms255 ms100%288
07Anthropic196 ms257 ms100%288
08Perplexity200 ms428 ms100%288
09reka207 ms449 ms100%288
10meta-llama218 ms377 ms100%288
11inference-net234 ms561 ms100%288
12Groq236 ms300 ms100%288
13xAI238 ms425 ms100%288
14writer246 ms294 ms100%288
15novita247 ms309 ms100%288
16Fireworks AI252 ms276 ms100%288
17replicate252 ms542 ms100%288
18Mistral255 ms344 ms100%288
19OpenAI257 ms589 ms100%288
20friendli269 ms372 ms100%288
21Together AI270 ms654 ms100%288
22siliconflow285 ms289 ms100%288
23targon317 ms329 ms100%288
24ai21326 ms467 ms100%288
25yi-01ai330 ms1,176 ms100%288
26kimi356 ms463 ms100%288
27baichuan359 ms390 ms100%288
28nebius366 ms374 ms100%288
29SambaNova371 ms399 ms100%288
30ernie408 ms451 ms100%288
31DeepSeek438 ms498 ms100%288
32aleph-alpha442 ms473 ms100%288
33nscale450 ms454 ms100%288
34minimax487 ms519 ms100%288
35deepinfra536 ms1,045 ms99.7%288
36glm537 ms566 ms100%288
37upstage606 ms646 ms100%288
38sarvam625 ms645 ms99.0%288
39sensenova632 ms1,504 ms99.7%288
40iflytek650 ms1,590 ms100%288
41stepfun658 ms1,597 ms100%288
42doubao694 ms1,595 ms99.7%288
43qwen713 ms772 ms100%288
44featherless757 ms3,089 ms100%288
45hyperbolic841 ms926 ms100%288
46hunyuan1,657 ms1,739 ms100%288
46 providersRegion report

US (Central)

RankProviderTypical responsep50 TTFB · msSlower responsesp95 TTFB · msAvailabilitySuccessful probesSamplesLast 24 hours
01Fireworks AI28 ms71 ms100%288
02baseten46 ms60 ms100%288
03meta-llama51 ms112 ms100%288
04Google59 ms149 ms100%288
05Cohere63 ms132 ms100%288
06OpenRouter64 ms142 ms100%288
07Cerebras76 ms149 ms100%288
08TypeSafe76 ms96 ms100%288
09siliconflow79 ms103 ms100%288
10novita80 ms160 ms100%288
11Perplexity87 ms143 ms100%288
12replicate88 ms147 ms100%288
13targon90 ms102 ms100%288
14Anthropic94 ms158 ms100%288
15reka96 ms183 ms99.7%288
16xAI116 ms186 ms100%288
17Groq117 ms190 ms100%288
18hyperbolic124 ms266 ms100%288
19Together AI126 ms211 ms99.7%288
20OpenAI147 ms226 ms100%288
21writer155 ms213 ms100%288
22inference-net160 ms435 ms100%288
23friendli163 ms241 ms100%288
24deepinfra164 ms643 ms100%288
25SambaNova170 ms215 ms100%288
26nebius175 ms209 ms100%288
27Mistral200 ms276 ms100%288
28yi-01ai220 ms258 ms100%288
29ai21235 ms372 ms99.7%288
30nscale236 ms242 ms100%288
31baichuan237 ms866 ms100%288
32glm253 ms296 ms100%288
33aleph-alpha256 ms313 ms100%288
34kimi268 ms403 ms100%288
35ernie269 ms282 ms100%288
36DeepSeek316 ms384 ms100%288
37minimax339 ms361 ms100%288
38upstage354 ms374 ms100%288
39featherless368 ms2,810 ms99.7%288
40stepfun398 ms1,006 ms100%288
41qwen406 ms455 ms100%288
42sensenova406 ms1,046 ms99.7%288
43iflytek410 ms1,128 ms100%288
44doubao457 ms1,086 ms100%288
45sarvam499 ms518 ms100%288
46hunyuan1,382 ms1,398 ms100%288
46 providersRegion report

Is any AI API down right now?

1 confirmed outage(s) in the last 24 hours: Google

46 providers · probed every 5 minutes from 4 regions · last probe Oct 22, 08:41 UTC

Status pages currently reporting an open incident: OpenAI (minor “ChatGPT conversion failures in India, Japan and Korea. And Response AP”) — self-reported by the provider, checked Oct 22, 08:17 UTC. Our probes test the API host; errors limited to individual models are visible only to the provider.

Live status per provider: OpenAI API · Claude API (Anthropic) · Gemini API (Google) · Mistral API · DeepSeek API · xAI API · Groq API · Cerebras API · SambaNova API · Together AI API · Fireworks AI API · OpenRouter API · TypeSafe API

Endpoint availability · hourly probe results

Last 72 hours · UTC · newest time on the right · current hour is incomplete.

One row per provider. Each block shows the results of one hour’s endpoint probes, combined across regions. Adjacent hours with all successes or no data are joined.

all probes succeededsome failed (under half)half or more failedno probes recorded

Probe outcomes per provider, hour by hour, last 72 hours (all regions combined)Each row is a provider. Blocks show the outcome of hourly endpoint probes across all regions: all succeeded, some failed, half or more failed, or no data. Text counts follow the chart; a failed probe alone does not establish a provider outage.Oct 20Oct 21Oct 22nowOct 19, 09:00 – Oct 21, 09:00 UTC: all 2358 probes succeededOct 21, 10:00 UTC: 2 of 48 probes failed — Europe (Germany) ([Errno 101] Network is unreachable)Oct 21, 11:00 – Oct 22, 08:00 UTC: all 1044 probes succeededOct 19, 09:00 – Oct 21, 09:00 UTC: all 2358 probes succeededOct 21, 10:00 UTC: 4 of 48 probes failed — Europe (Germany) ([Errno 101] Network is unreachable)Oct 21, 11:00 – Oct 22, 08:00 UTC: all 1044 probes succeededOct 19, 09:00 – Oct 22, 08:00 UTC: all 3450 probes succeededOct 19, 09:00 – Oct 21, 09:00 UTC: all 2358 probes succeededOct 21, 10:00 UTC: 2 of 48 probes failed — Europe (Germany) ([Errno 101] Network is unreachable)Oct 21, 11:00 – Oct 22, 08:00 UTC: all 1044 probes succeededOct 19, 09:00 – Oct 22, 08:00 UTC: all 3450 probes succeededOct 19, 09:00 – Oct 21, 09:00 UTC: all 2358 probes succeededOct 21, 10:00 UTC: 2 of 48 probes failed — Europe (Germany) ([Errno 101] Network is unreachable)Oct 21, 11:00 – Oct 22, 08:00 UTC: all 1044 probes succeededOct 19, 09:00 – Oct 21, 09:00 UTC: all 2358 probes succeededOct 21, 10:00 UTC: 2 of 48 probes failed — Europe (Germany) ([Errno 101] Network is unreachable)Oct 21, 11:00 – Oct 22, 08:00 UTC: all 1044 probes succeededOct 19, 09:00 – Oct 21, 09:00 UTC: all 2358 probes succeededOct 21, 10:00 UTC: 2 of 48 probes failed — Europe (Germany) ([Errno 101] Network is unreachable)Oct 21, 11:00 – Oct 22, 05:00 UTC: all 912 probes succeededOct 22, 06:00 UTC: 1 of 48 probes failed — Asia (Tokyo) (timeout)Oct 22, 07:00 – Oct 22, 08:00 UTC: all 84 probes succeededOct 19, 09:00 – Oct 20, 03:00 UTC: all 912 probes succeededOct 20, 04:00 UTC: 1 of 48 probes failed — Asia (Tokyo) (timeout)Oct 20, 05:00 – Oct 22, 08:00 UTC: all 2490 probes succeededOct 19, 09:00 – Oct 21, 09:00 UTC: all 2358 probes succeededOct 21, 10:00 UTC: 2 of 48 probes failed — Europe (Germany) ([Errno 101] Network is unreachable)Oct 21, 11:00 – Oct 22, 06:00 UTC: all 960 probes succeededOct 22, 07:00 UTC: 1 of 48 probes failed — US (Central) (http 502)Oct 22, 08:00 UTC: all 36 probes succeededOct 19, 09:00 – Oct 19, 10:00 UTC: all 96 probes succeededOct 19, 11:00 UTC: 1 of 48 probes failed — Europe (Germany) (timeout)Oct 19, 12:00 – Oct 20, 14:00 UTC: all 1299 probes succeededOct 20, 15:00 UTC: 1 of 48 probes failed — Europe (Germany) ([Errno 101] Network is unreachable)Oct 20, 16:00 – Oct 21, 09:00 UTC: all 867 probes succeededOct 21, 10:00 UTC: 1 of 48 probes failed — Europe (Germany) ([Errno 101] Network is unreachable)Oct 21, 11:00 – Oct 22, 08:00 UTC: all 1044 probes succeededOct 19, 09:00 – Oct 22, 08:00 UTC: all 3450 probes succeededOct 19, 09:00 – Oct 21, 09:00 UTC: all 2358 probes succeededOct 21, 10:00 UTC: 1 of 48 probes failed — Europe (Germany) ([Errno 101] Network is unreachable)Oct 21, 11:00 – Oct 22, 08:00 UTC: all 1044 probes succeeded
openaianthropicgooglemistraldeepseekxaigroqcerebrassambanovatogetherfireworksopenroutertypesafe
Probe counts by provider
OpenAI: 2 failed / 3,450 endpoint probesAnthropic: 4 failed / 3,450 endpoint probesGoogle: 0 failed / 3,450 endpoint probesMistral: 2 failed / 3,450 endpoint probesDeepSeek: 0 failed / 3,450 endpoint probesxAI: 2 failed / 3,450 endpoint probesGroq: 2 failed / 3,450 endpoint probesCerebras: 3 failed / 3,450 endpoint probesSambaNova: 1 failed / 3,450 endpoint probesTogether AI: 3 failed / 3,450 endpoint probesFireworks AI: 3 failed / 3,450 endpoint probesOpenRouter: 0 failed / 3,450 endpoint probesTypeSafe: 1 failed / 3,450 endpoint probes

Full incident log

Fastest by region

Lowest typical response (p50 TTFB) · last 24 hours

Methodology
Asia (Tokyo)Fireworks AI19msEurope (Germany)Fireworks AI98msSouth America (São Paulo)OpenRouter62msUS (Central)Fireworks AI28ms

Which AI API is fastest right now?

As of October 22, 2026, the fastest-responding AI inference API by edge latency (time-to-first-byte) is fireworks at 19 ms p50, measured from Asia (Tokyo) (100% uptime, n=288). Rankings differ by region.

Edge latency (TTFB) measures the network and API endpoint; inference TTFT is measured separately. Methodology

All 46 providers
ai21aleph-alphaAnthropicbaichuanbasetenCerebrasCoheredeepinfraDeepSeekdoubaoerniefeatherlessFireworks AIfriendliglmGoogleGroqhunyuanhyperboliciflytekinference-netkimimeta-llamaminimaxMistralnebiusnovitanscaleOpenAIOpenRouterPerplexityqwenrekareplicateSambaNovasarvamsensenovasiliconflowstepfuntargonTogether AITypeSafeupstagewriterxAIyi-01ai

Provider comparisons