Benchmark report
Is the Claude API slow?
Opus 5 time to first token from 6 cities, measured daily
Consistent worldwide
First visible token arrived in 1043 ms from Montreal and 1251 ms from Mumbai. All test locations waited about the same.
- Fastest
- 1043 ms
- Montreal
- Slowest
- 1251 ms
- Mumbai
- Successful requests
- 100%
- 6 locations
Response time around the world
Montreal1043 ms
San Francisco1073 ms
Amsterdam1105 ms
Tokyo1158 ms
Singapore1205 ms
Mumbai1251 ms
One block of waiting before the stream opens
The stream opened at the same moment the first visible token arrived, within 7 ms in every city, and the whole reply to our one-word prompt was done 1 to 53 ms later. Anthropic streamed no thinking phase for this prompt in almost every request (one of 20 in Singapore carried one, at about a second), so the wait shows up as one block before the stream opens. Connecting took 2 to 9 ms and the secure handshake 4 to 12 ms; the rest is queueing plus low-effort thinking on Anthropic's side. In the controlled run of 27 September that block took 1.2 seconds in San Francisco and Tokyo and 1.5 seconds in Singapore.
Click the location to see each stage.
Singapore1538 ms
Singapore- Finding the server
- 3 ms
- Reaching the server
- 3 ms
- Setting up security
- 6 ms
- Waiting for the server: most of the time
- 1526 ms
- Receiving the response
- 3 ms
- Total
- 1538 ms
- Response opened
- 1478 ms
- First visible word
- 1485 ms
- Answer complete
- 1538 ms
Finding the server is measured once per location, so it sits outside these bars. Open a city to see it.
Bars show the typical request, so the parts add up to its total. Why
Compare all locationsHide the comparison
Click a city to see its stage-by-stage breakdown.
Amsterdam1302 ms
Amsterdam- Finding the server
- 1 ms
- Reaching the server
- 9 ms
- Setting up security
- 12 ms
- Waiting for the server: most of the time
- 1269 ms
- Receiving the response
- 12 ms
- Total
- 1302 ms
- Response opened
- 1290 ms
- First visible word
- 1291 ms
- Answer complete
- 1302 ms
San Francisco1201 ms
San Francisco- Finding the server
- 15 ms
- Reaching the server
- 3 ms
- Setting up security
- 6 ms
- Waiting for the server: most of the time
- 1191 ms
- Receiving the response
- 1 ms
- Total
- 1201 ms
- Response opened
- 1200 ms
- First visible word
- 1200 ms
- Answer complete
- 1201 ms
Montreal1291 ms
Montreal- Finding the server
- 2 ms
- Reaching the server
- 3 ms
- Setting up security
- 4 ms
- Waiting for the server: most of the time
- 1276 ms
- Receiving the response
- 8 ms
- Total
- 1291 ms
- Response opened
- 1283 ms
- First visible word
- 1284 ms
- Answer complete
- 1291 ms
Singapore1538 ms
Singapore- Finding the server
- 3 ms
- Reaching the server
- 3 ms
- Setting up security
- 6 ms
- Waiting for the server: most of the time
- 1526 ms
- Receiving the response
- 3 ms
- Total
- 1538 ms
- Response opened
- 1478 ms
- First visible word
- 1485 ms
- Answer complete
- 1538 ms
Tokyo1214 ms
Tokyo- Finding the server
- 1 ms
- Reaching the server
- 2 ms
- Setting up security
- 4 ms
- Waiting for the server: most of the time
- 1202 ms
- Receiving the response
- 6 ms
- Total
- 1214 ms
- Response opened
- 1208 ms
- First visible word
- 1208 ms
- Answer complete
- 1214 ms
Mumbai1311 ms
Mumbai- Finding the server
- 2 ms
- Reaching the server
- 3 ms
- Setting up security
- 7 ms
- Waiting for the server: most of the time
- 1301 ms
- Receiving the response
- 0 ms
- Total
- 1311 ms
- Response opened
- 1306 ms
- First visible word
- 1306 ms
- Answer complete
- 1311 ms
The phases come from the controlled run of 27 September 2026. Typical and slower times pool the daily requests. 6 of 6 test locations included: Amsterdam, San Francisco, Montreal, Singapore, Tokyo, Mumbai.
Finding the server is measured once per location, so it sits outside these bars. Open a city to see it.
Bars show the typical request, so the parts add up to its total. Why
Anthropic still feels slow?
- Check Anthropic’s status page
- Elevated latency and error rates are posted there. A slow first token during an incident is not your region.
- The day matters more than the city
- Every city resolved the API to one address, and the six cities sat 20% apart on typical first token while every city's daily median ranged from about 0.8 to 2.1 seconds across the window. That address does not say where the model runs.
Response time over the last 29 days
A dot marks a day well above that location’s usual level.
Show the numbers
| Day | Amsterdam | San Francisco | Montreal | Singapore | Tokyo | Mumbai |
|---|---|---|---|---|---|---|
| 11 September 2026 | 1844 ms | 1218 ms | 1221 ms | 1362 ms | 1276 ms | 1357 ms |
| 12 September 2026 | 1402 ms | 1373 ms | 1183 ms | 1426 ms | 1427 ms | 1494 ms |
| 13 September 2026 | 1408 ms | 1356 ms | 1362 ms | 1479 ms | 1434 ms | 1515 ms |
| 14 September 2026 | 1323 ms | 1265 ms | 1314 ms | 1519 ms | 1457 ms | 1530 ms |
| 15 September 2026 | 1366 ms | 1311 ms | 1235 ms | 1501 ms | 1399 ms | 1486 ms |
| 16 September 2026 | 1297 ms | 1225 ms | 1179 ms | 1423 ms | 1318 ms | 1455 ms |
| 17 September 2026 | 1793 ms | 1641 ms | 1639 ms | 2077 ms | 1855 ms | 1911 ms |
| 18 September 2026 | 1810 ms | 1657 ms | 1671 ms | 1902 ms | 1812 ms | 1975 ms |
| 19 September 2026 | 1272 ms | 1261 ms | 1226 ms | 1504 ms | 1327 ms | 1463 ms |
| 20 September 2026 | 1155 ms | 1157 ms | 1231 ms | 1408 ms | 1453 ms | 1506 ms |
| 21 September 2026 | 1756 ms | 1802 ms | 1751 ms | 1792 ms | 1739 ms | 1878 ms |
| 22 September 2026 | 1724 ms | 1780 ms | 1411 ms | 1705 ms | 4740 ms ▲ | 1643 ms |
| 23 September 2026 | 1288 ms | 1273 ms | 1194 ms | 1500 ms | 1367 ms | 1450 ms |
| 24 September 2026 | 1244 ms | 1259 ms | 1131 ms | 1343 ms | 1258 ms | 1419 ms |
| 25 September 2026 | 1211 ms | 1275 ms | 1338 ms | 1367 ms | 1281 ms | 1476 ms |
| 26 September 2026 | 1166 ms | 1130 ms | 1081 ms | 1290 ms | 1179 ms | 1281 ms |
| 27 September 2026 | 1090 ms | 1016 ms | 1021 ms | 1205 ms | 1127 ms | 1212 ms |
| 28 September 2026 | 1335 ms | 1283 ms | 1194 ms | 1466 ms | 1285 ms | 1429 ms |
| 29 September 2026 | 1390 ms | 1278 ms | 1229 ms | 1662 ms | 1471 ms | 1479 ms |
| 30 September 2026 | 1437 ms | 1495 ms | 1329 ms | 2072 ms | 1366 ms | 1393 ms |
| 1 October 2026 | 1105 ms | 2038 ms | 967 ms | 1265 ms | 1099 ms | 1200 ms |
| 2 October 2026 | 1121 ms | 1055 ms | 1043 ms | 1175 ms | 1152 ms | 1251 ms |
| 3 October 2026 | 1051 ms | 969 ms | 989 ms | 1137 ms | 1058 ms | 1190 ms |
| 4 October 2026 | 1044 ms | 1013 ms | 954 ms | 1146 ms | 1072 ms | 1162 ms |
| 5 October 2026 | 1024 ms | 811 ms | 790 ms | 974 ms | 1092 ms | 1044 ms |
| 6 October 2026 | 1292 ms | 1210 ms | 1166 ms | 1428 ms | 1389 ms | 1606 ms |
| 7 October 2026 | 1105 ms | 1081 ms | 1020 ms | 1265 ms | 1189 ms | 1269 ms |
| 9 October 2026 | 1028 ms | 974 ms | 978 ms | 1113 ms | 1196 ms | 1345 ms |
| 10 October 2026 | 885 ms | 1689 ms | 1797 ms | 1125 ms | 1094 ms | 1190 ms |
About this measurement
POST api.anthropic.com/v1/messages · 65 requests per location over 13 daily runs · 27 September 2026 to 10 October 2026
What we tested · AI response. Time to first visible token from claude-opus-5 on a one-word prompt: streaming, effort low with adaptive thinking on, 256 max tokens, paid key.
How LatencyRadar measures response time →
Technical details
Doesn’t measure: The Claude app, tokens per second, long prompts, other Claude models, or the Batch API.
One Anthropic surface: time to first visible token from claude-opus-5 at api.anthropic.com/v1/messages, a one-word prompt, streaming, effort low, 256 max tokens, paid key. On Opus 5 thinking is adaptive when the request does not disable it, so the wait includes a short thinking phase that is not streamed. That matches the low-effort setting used for the ChatGPT and Grok surfaces, although each model can spend that effort differently.
The published measurements cover 27 September to 10 October 2026. Typical first-token times are 1.04 to 1.25 seconds across six cities, Montreal quickest and Mumbai slowest. Some days were slower everywhere: from 28 to 30 September every city's daily median ran at 1.2 to 2.1 seconds, and on 6 October at 1.2 to 1.6 seconds. Other slow days hit one or two cities: San Francisco took 2.0 seconds on 1 October, and on 10 October San Francisco and Montreal took 1.7 and 1.8 seconds while the other cities stayed between 0.9 and 1.2 seconds. Check both the typical and the slower times when deciding how long your app should show a waiting state.
| Test location | Typical response time | Slower response time | Requests |
|---|---|---|---|
| Amsterdam | 1105 ms | 1702 ms | 65 |
| San Francisco | 1073 ms | 2027 ms | 65 |
| Montreal | 1043 ms | 1857 ms | 65 |
| Singapore | 1205 ms | 2072 ms | 65 |
| Tokyo | 1158 ms | 1731 ms | 65 |
| Mumbai | 1251 ms | 1665 ms | 65 |
Typical: half of the requests finished within this time (technical: p50 of time to first visible token). Slower: 95% of requests finished within this time (technical: p95). Statistics
- Request
- Authenticated streaming POST to api.anthropic.com/v1/messages
- Measured from
- Amsterdam · San Francisco · Montreal · Singapore · Tokyo · Mumbai
- Requests
- Daily measurements from 27 September 2026 to 10 October 2026. 6 of 6 test locations included: Amsterdam, San Francisco, Montreal, Singapore, Tokyo, Mumbai. · 27 September 2026 to 10 October 2026
- Model
- claude-opus-5 · output_config effort "low" (adaptive thinking on, not streamed) · max_tokens 256 · anthropic-version 2023-06-01
- Prompt
- "Say 'ok' and nothing else." Identical across every city and every LLM API we benchmark
- Comparison figures
- ChatGPT (gpt-5.6-sol, reasoning_effort low) and Grok (grok-4.6, reasoning_effort low) come from the same daily collection between 27 September and 10 October 2026
- Timings taken
- Parsed from the token stream: stream open, first visible token, completion, plus DNS, connect and TLS
- Report coverage
- 6 of 6 test locations included: Amsterdam, San Francisco, Montreal, Singapore, Tokyo, Mumbai.
- Coverage changes
- Mumbai (measured since 7 September 2026) joined this report on 21 September 2026. Earlier daily bands keep their previous location set.
- Daily requests in each location
- Amsterdam: 65 · San Francisco: 65 · Montreal: 65 · Singapore: 65 · Tokyo: 65 · Mumbai: 65
- Typical response time
- Median of the eligible daily requests in each included location.
- Slower response time
- The 95th percentile of the same eligible daily requests in each included location.
Claude API
https://api.anthropic.com/v1/messages
Independent measurement by LatencyRadar. Not affiliated with Anthropic. · How we measure (v1.3)
How fast is your API compared to Anthropic's?
Run a free speed test from multiple cities and find out where your users are waiting. No setup, no account required.
No account required · Takes about 30 seconds.
