AI API latency report · Oct 2026

ChatGPT vs Claude vs Grok, who's fastest?

Across six cities and 1,170 streaming requests, median time to first visible token was 990 ms for ChatGPT, 1.1 s for Claude and 2.6 s for Grok.

What the data shows

  1. Finding 1

    ChatGPT was fastest overall

    990 ms median time to first token, ahead of Claude at 1.1 s and Grok at 2.6 s.

    • 67/78 city-days vs Claude
    • 78/78 vs Grok
    See the day-by-day chart →

  2. Finding 2

    Provider choice mattered more than location

    In every city, the gap between providers was larger than the gap between locations.

    • Provider gap: ≥1,390 ms
    • City gap: ≤578 ms
    See the table by city →

  3. Finding 3

    Most of the delay was not network

    Network setup took just 12–19 ms; most of the wait came before the first visible token.

    • Network: 12–19 ms
    • First token: 887–2,750 ms
    See where the time goes →

Time to first token, day by day

Daily median time to first token, 11 September 2026 to 10 October 2026. Each point is the median of the compared cities’ daily medians, 5 requests per city.
01000200030004000 ms11 Sept10 Oct30711158953
Show the numbers
DayChatGPTClaudeGrok
11 September 20261319 ms1317 ms2874 ms
12 September 20261124 ms1414 ms1837 ms
13 September 20261211 ms1421 ms2938 ms
14 September 20261403 ms1390 ms2519 ms
15 September 20261336 ms1383 ms2573 ms
16 September 20261207 ms1308 ms2633 ms
17 September 20261000 ms1824 ms2518 ms
18 September 20261393 ms1811 ms2961 ms
19 September 20261089 ms1300 ms2828 ms
20 September 2026997 ms1320 ms2456 ms
21 September 2026918 ms1774 ms2792 ms
22 September 2026962 ms1715 ms2616 ms
23 September 20261192 ms1328 ms2485 ms
24 September 20261020 ms1259 ms2636 ms
25 September 20261111 ms1310 ms1445 ms
26 September 2026729 ms1173 ms1931 ms
27 September 2026896 ms1109 ms2390 ms
28 September 2026825 ms1310 ms2810 ms
29 September 20261080 ms1431 ms2542 ms
30 September 20261067 ms1415 ms2600 ms
1 October 20261180 ms1153 ms2335 ms
2 October 2026964 ms1137 ms2444 ms
3 October 2026775 ms1055 ms2874 ms
4 October 20261004 ms1058 ms2689 ms
5 October 2026883 ms999 ms2692 ms
6 October 20261107 ms1341 ms3095 ms
7 October 20261002 ms1147 ms2504 ms
9 October 20261165 ms1071 ms2919 ms
10 October 2026953 ms1158 ms3071 ms

Does your location matter?

For AI APIs, your provider matters more than your server location.

Typical time to first token by city: the median of every daily request from that city over the window.
01000200030004000 msChatGPT from Amsterdam: 1025 ms typical time to first token; slower requests 1612 ms.Claude from Amsterdam: 1105 ms typical time to first token; slower requests 1702 ms.Grok from Amsterdam: 3026 ms typical time to first token; slower requests 4975 ms.AmsterdamChatGPT from San Francisco: 928 ms typical time to first token; slower requests 1517 ms.Claude from San Francisco: 1073 ms typical time to first token; slower requests 2027 ms.Grok from San Francisco: 2474 ms typical time to first token; slower requests 3716 ms.San FranciscoChatGPT from Montreal: 748 ms typical time to first token; slower requests 1424 ms.Claude from Montreal: 1043 ms typical time to first token; slower requests 1857 ms.Grok from Montreal: 2591 ms typical time to first token; slower requests 4516 ms.MontrealChatGPT from Singapore: 977 ms typical time to first token; slower requests 1963 ms.Claude from Singapore: 1205 ms typical time to first token; slower requests 2072 ms.Grok from Singapore: 2656 ms typical time to first token; slower requests 4910 ms.SingaporeChatGPT from Tokyo: 1003 ms typical time to first token; slower requests 1993 ms.Claude from Tokyo: 1158 ms typical time to first token; slower requests 1731 ms.Grok from Tokyo: 2744 ms typical time to first token; slower requests 4845 ms.TokyoChatGPT from Mumbai: 1326 ms typical time to first token; slower requests 2316 ms.Claude from Mumbai: 1251 ms typical time to first token; slower requests 1665 ms.Grok from Mumbai: 2641 ms typical time to first token; slower requests 5010 ms.Mumbai
  • ChatGPT
  • Claude
  • Grok
Show exact values
CityChatGPTClaudeGrok
Amsterdam1025 msslower 1.6 s1105 msslower 1.7 s3026 msslower 5.0 s
San Francisco928 msslower 1.5 s1073 msslower 2.0 s2474 msslower 3.7 s
Montreal748 msslower 1.4 s1043 msslower 1.9 s2591 msslower 4.5 s
Singapore977 msslower 2.0 s1205 msslower 2.1 s2656 msslower 4.9 s
Tokyo1003 msslower 2.0 s1158 msslower 1.7 s2744 msslower 4.8 s
Mumbai1326 msslower 2.3 s1251 msslower 1.7 s2641 msslower 5.0 s

Bold is the lowest typical time in that city. "Slower" is the time 95% of the daily requests from that city finished within.

Where does the wait happen?

How much of the wait a closer server could remove, and how much is the provider’s own queue and model.

Both bars share one scale. They come from one test of 20 requests per city on 27 September 2026, not from the two weeks of daily requests behind the typical times, so they won't match those exactly.
  • ChatGPT
    Network setup
    19 ms
    Waiting for the model
    868 ms

    First visible token after 887 ms in total.

  • Claude
    Network setup
    12 ms
    Waiting for the model
    1276 ms

    First visible token after 1288 ms in total.

  • Grok
    Network setup
    18 ms
    Waiting for the model
    2732 ms

    First visible token after 2750 ms in total.

Network setup is finding the server, reaching it and setting up a secure connection. Waiting for the model is everything after that until the first visible token: queueing and the model’s own work.

ChatGPT vs Claude vs Grok

27 September – 10 October 2026 · six test locations

MetricChatGPTClaudeGrok
Typical TTFT990 ms1132 ms2649 ms
Location spread77%20%22%

ChatGPT uses gpt-5.6-sol; Claude uses claude-opus-5; Grok uses grok-4.6. Methodology →

Show full benchmark details ↓
DetailChatGPTClaudeGrok
Modelgpt-5.6-solclaude-opus-5grok-4.6
Slower requests, p95 (median city)1788 ms1794 ms4878 ms
Fastest locationMontreal · 748 msMontreal · 1043 msSan Francisco · 2474 ms
Slowest locationMumbai · 1326 msMumbai · 1251 msAmsterdam · 3026 ms
Days when every city slowed together000
Requests in this window390 over 13 daily runs390 over 13 daily runs390 over 13 daily runs
Latest controlled run27 September 202627 September 202627 September 2026

How we test

We send the same short streaming prompt to each provider from six cities and measure time to first visible token. Typical and slower (p95) times both come from the same daily requests. Only the split between network setup and waiting for the model comes from a separate controlled run.

Claude Opus 5 includes its adaptive thinking phase; ChatGPT and Grok run at their lowest reasoning effort.

Full testing methodology →

How does your API compare?

Test your endpoint from the same 6 cities used here and see your response time next to these numbers.

Run a free latency test

No account required · Takes about 30 seconds.

Explore each API