Is the Grok API slow? grok-4.6 time to first token from 5 cities, measured daily
We ran real, authenticated streaming completions against xAI's grok-4.6, with reasoning effort set to "low", from Amsterdam, San Francisco, Montreal, Singapore and Tokyo, every day from 27 August to 4 September 2026, plus two controlled runs of 20 requests per city, parsing the token stream ourselves and timestamping the moment the first visible token arrived.
What we tested · AI response
Time to first visible token from grok-4.6 on a one-word prompt: streaming, reasoning_effort low, 256 max tokens, paid key.
Doesn’t measure: Grok inside X or at grok.com, tokens per second, long prompts or other xAI models.
Quick take
Grok's first visible token typically arrived in 2.45 to 2.73 seconds, measured 40 times per city over eight days, an 11% spread, so where you call from barely matters. The stream opens in 0.13 to 0.29 seconds and reasoning deltas start about half a second in; the rest is the model thinking. Plan on two and a half seconds before the first visible word and 3.3 to 3.9 seconds for slower responses. ChatGPT typically took 1.0 second and Claude 1.4 in the same daily collection.
Where in the world is it fast?
Montreal was quickest at 2.45 s to the first visible token, then San Francisco (2.57 s), Amsterdam (2.57 s), Tokyo (2.69 s) and Singapore (2.73 s). A 275 ms spread on a two-and-a-half-second wait means geography is nearly irrelevant for grok-4.6: when the model spends most of the time thinking, moving closer to the servers buys you very little. The day matters more than the city. Tokyo's daily median ranged from 1.2 to 3.2 s across the eight runs.
Amsterdam2572 msslower 3565 ms
San Francisco2565 msslower 3809 ms
Montreal2453 msslower 3357 ms
Singapore2728 msslower 3920 ms
Tokyo2691 msslower 3332 ms
If xAI is slow for you
Check xAI’s status page- What the number includes
- Model queueing and low-effort reasoning. The stream opens early, so a fast stream open does not mean a fast first word.
- Regions
- The request starts at our test location. Network evidence does not tell us where xAI runs the model.
- Comparison basis
- Same prompt and low effort as the ChatGPT and Claude surfaces, collected on the same days.
Where is the time actually going?
Measured from the stream itself, one grok-4.6 call has three acts: the stream opens in 0.13–0.29 s (that part is network and edge), reasoning deltas start arriving around 0.5–0.65 s, and the first visible token lands at 2.1–2.6 s in the latest controlled run, with the trivial answer completing within a few milliseconds after it. Unlike the ChatGPT and Claude APIs, which hold the response until the first token is ready, Grok's wait happens inside an open stream. The first visible word is what a user sees appear, so that is the headline number here. The transport phases below are supporting detail, and the long "download" phase is the in-stream reasoning, not bytes on the wire.
Click a city to see its stage-by-stage breakdown.
Amsterdam2612 ms
Amsterdam- Finding the server
- 1 ms
- Reaching the server
- 10 ms
- Setting up security
- 22 ms
- Waiting for the server
- 381 ms
- Receiving the response: most of the time
- 2199 ms
- Total
- 2612 ms
- Response opened
- 281 ms
- First visible word
- 2609 ms
- Answer complete
- 2612 ms
San Francisco2131 ms
San Francisco- Finding the server
- 15 ms
- Reaching the server
- 4 ms
- Setting up security
- 8 ms
- Waiting for the server
- 99 ms
- Receiving the response: most of the time
- 2020 ms
- Total
- 2131 ms
- Response opened
- 128 ms
- First visible word
- 2130 ms
- Answer complete
- 2131 ms
Montreal2556 ms
Montreal- Finding the server
- 1 ms
- Reaching the server
- 22 ms
- Setting up security
- 29 ms
- Waiting for the server
- 156 ms
- Receiving the response: most of the time
- 2349 ms
- Total
- 2556 ms
- Response opened
- 210 ms
- First visible word
- 2555 ms
- Answer complete
- 2556 ms
Singapore2477 ms
Singapore- Finding the server
- 6 ms
- Reaching the server
- 3 ms
- Setting up security
- 8 ms
- Waiting for the server
- 253 ms
- Receiving the response: most of the time
- 2213 ms
- Total
- 2477 ms
- Response opened
- 286 ms
- First visible word
- 2474 ms
- Answer complete
- 2477 ms
Tokyo2243 ms
Tokyo- Finding the server
- 4 ms
- Reaching the server
- 3 ms
- Setting up security
- 6 ms
- Waiting for the server
- 190 ms
- Receiving the response: most of the time
- 2044 ms
- Total
- 2243 ms
- Response opened
- 201 ms
- First visible word
- 2242 ms
- Answer complete
- 2243 ms
The bars and phases below come from the controlled run of 4 September 2026 (20 requests per city), so they add up within that run. The typical and slower times in the table above pool every daily run, which is why the two sets of numbers differ.
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
Every city, side by side
One row per city, with the typical time highlighted. For the stage-by-stage detail behind any city, open it in the chart above.
Grok API · 27 August 2026 – 4 September 2026
| Test location | Typical first word | On a bad day | Slowest seen |
|---|---|---|---|
Amsterdam | 2572 ms | 3565 ms | 3105 ms |
San Francisco | 2565 ms | 3809 ms | 4406 ms |
Montreal Fastest | 2453 ms | 3357 ms | 3999 ms |
Singapore Slowest | 2728 ms | 3920 ms | 4000 ms |
Tokyo | 2691 ms | 3332 ms | 3733 ms |
How it has changed over time
One panel per city, one point per run, all panels on the same scale. Daily runs and the controlled 20-request runs are shown as separate charts because they measure different things.
Time to first visible token from grok-4.6 on a one-word prompt: streaming, reasoning_effort low, 256 max tokens, paid key.
Daily response time
Daily median of 5 requests, per city, 8 August 2026 to 7 September 2026. Open markers on the baseline are runs with fewer than 3 valid requests; they carry no value. A dashed line marks the day a city joined the measurements.
Collecting since 7 Sept
Show the numbers
| Run | Amsterdam | San Francisco | Montreal | Singapore | Tokyo | Mumbai |
|---|---|---|---|---|---|---|
| 27 August 2026 | 2431 ms | 1988 ms | 2331 ms | 2874 ms | 2992 ms | — |
| 28 August 2026 | 2451 ms | 3093 ms | 2038 ms | 2596 ms | 1218 ms | — |
| 29 August 2026 | 3216 ms | 2259 ms | 2841 ms | 2728 ms | 2405 ms | — |
| 31 August 2026 | 3506 ms | 2390 ms | 2016 ms | 3061 ms | 3150 ms | — |
| 1 September 2026 | 2703 ms | 2570 ms | 3289 ms | 2643 ms | 2675 ms | — |
| 2 September 2026 | 2572 ms | 2669 ms | 2419 ms | 2856 ms | 2877 ms | — |
| 3 September 2026 | 2170 ms | 2198 ms | 2383 ms | 2421 ms | 2108 ms | — |
| 4 September 2026 | 3439 ms | 3345 ms | 2502 ms | 2678 ms | 2998 ms | — |
| 5 September 2026 | 2883 ms | 1936 ms | 2161 ms | 3742 ms | 2758 ms | — |
| 6 September 2026 | 2653 ms | 2323 ms | 2311 ms | 2863 ms | 2524 ms | — |
| 7 September 2026 | 2507 ms | 2363 ms | 2608 ms | 3066 ms | 2059 ms | 3001 ms |
Slower response time in the controlled runs
Slower response time in each 20-request run, per city, 9 June 2026 to 7 September 2026. Open markers on the baseline are runs with fewer than 15 valid requests; they carry no value. A dashed line marks the day a city joined the measurements.
Collecting since 27 Aug
Collecting since 27 Aug
Collecting since 27 Aug
Collecting since 27 Aug
Collecting since 27 Aug
Technical note: 95% of that run's requests finished within this time (technical: p95 of 20 requests).
Show the numbers
| Run | Amsterdam | San Francisco | Montreal | Singapore | Tokyo | Mumbai |
|---|---|---|---|---|---|---|
| 30 August 2026 | 4087 ms | 3531 ms | 3228 ms | 4073 ms | 3260 ms | — |
| 4 September 2026 | 3043 ms | 4087 ms | 3486 ms | 3766 ms | 3403 ms | — |
Global presence
Verdict
Uneven around the world
More than one server answers, but some cities wait noticeably longer, so some visitors may be reaching a distant one.
Based on 2 server addresses and a 158 ms spread between cities in the latest controlled run.
What these xAI results mean
One xAI surface: time to first visible token from grok-4.6 at api.x.ai/v1/chat/completions, a one-word prompt, streaming, reasoning_effort low, 256 max tokens, paid key. xAI opens the stream before the first token arrives, so stream-open time and first-token time differ here more than for OpenAI or Anthropic. We report the first visible token as the measurement, never the stream opening.
Not measured: Grok inside X or at grok.com, tokens per second, long prompts, other xAI models, image or voice.
In the published measurements, typical first-token times range from about 2.45 to 2.73 seconds across five cities. The stream can open before visible text arrives, so an early connection alone would make the wait appear shorter than it is. The first-token measurement captures the delay a user sees before the answer starts.
How this test was run
8 daily quick runs from 27 August 2026 to 4 September 2026 (40 requests per city), plus 2 controlled runs of 20 requests per city, the latest on 4 September 2026 · 27 August 2026 – 4 September 2026
Technical details
- Request
- Authenticated streaming POST to api.x.ai/v1/chat/completions
- Measured from
- Amsterdam · San Francisco · Montreal · Singapore · Tokyo
- Model
- grok-4.6 · reasoning effort "low" · max_tokens 256 · default temperature
- Prompt
- "Say 'ok' and nothing else." Identical across every city and every LLM API we benchmark
- Reasoning stream
- xAI reasons inside the open stream, so we also timestamp the first reasoning delta; it never counts as the first visible token
- Comparison figures
- ChatGPT (gpt-5.6-sol, reasoning_effort low) and Claude (claude-opus-5, effort low) come from the same daily collection over the same days
- Timings taken
- Parsed from the token stream: stream open, first visible token, completion, plus DNS, connect and TLS
- Typical response time
- Median of all 40 daily requests per city between 27 August 2026 and 4 September 2026; not an average of daily medians
- Slower response time
- Median of the 95th percentiles of the 2 most recent controlled 20-request runs per city
- Day-to-day range (daily medians)
- Amsterdam 2170–3506 ms · San Francisco 1988–3345 ms · Montreal 2016–3289 ms · Singapore 2421–3061 ms · Tokyo 1218–3150 ms
Grok API
https://api.x.ai/v1/chat/completions
Independent measurement by LatencyRadar. Not affiliated with xAI. · How we measure (v1.0)
Get the monthly AI latency report, by test location
One email a month: how fast the major AI providers respond, measured globally.
Check your inbox for the confirmation email. Nothing is sent until you confirm. Unsubscribe any time.
How fast is your API compared to xAI's?
Run a free speed test from multiple cities and find out where your users are waiting. No setup, no account required.
Takes about 30 seconds.