Benchmark report
Is the DeepSeek API slow?
DeepSeek V4 Flash time to first token from 6 cities, measured daily
Uneven around the world
First visible token typically arrived in 527 ms from Singapore and 734 ms from Montreal, with every city under 0.75 seconds. Asian cities wait less than North American and European ones, which suggests the model is served from Asia.
- Fastest
- 527 ms
- Singapore
- Slowest
- 734 ms
- Montreal
- Successful requests
- 100%
- 6 locations
Response time around the world
Singapore527 ms
Mumbai589 ms
Tokyo622 ms
Amsterdam700 ms
San Francisco728 ms
Montreal734 ms
The stream opens before the first token does
In the controlled run of 27 September DeepSeek opened the stream 317 to 536 ms before the first visible token arrived: 119 ms against 436 ms in Singapore, 283 ms against 819 ms in Amsterdam. OpenAI, Anthropic and Mistral hold the response until the first token is ready, so a client that times the first byte will read DeepSeek as faster than it is. Connecting took 2 to 6 ms and the secure handshake 3 to 14 ms; the wait before the stream opened was 119 ms in Singapore, 179 in Mumbai, 196 in Tokyo, 283 in Amsterdam, 296 in San Francisco and 339 in Montreal. Every city reached the same CloudFront address through a local point of presence, so that wait is the trip from the edge to DeepSeek, and it is shortest from Asia.
Click the location to see each stage.
Singapore504 ms
Singapore- Finding the server
- 3 ms
- Reaching the server
- 3 ms
- Setting up security
- 5 ms
- Waiting for the server
- 100 ms
- Receiving the response: most of the time
- 396 ms
- Total
- 504 ms
- Response opened
- 119 ms
- First visible word
- 436 ms
- Answer complete
- 504 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.
Amsterdam898 ms
Amsterdam- Finding the server
- 1 ms
- Reaching the server
- 6 ms
- Setting up security
- 5 ms
- Waiting for the server
- 272 ms
- Receiving the response: most of the time
- 615 ms
- Total
- 898 ms
- Response opened
- 283 ms
- First visible word
- 819 ms
- Answer complete
- 898 ms
San Francisco814 ms
San Francisco- Finding the server
- 17 ms
- Reaching the server
- 4 ms
- Setting up security
- 5 ms
- Waiting for the server
- 312 ms
- Receiving the response: most of the time
- 493 ms
- Total
- 814 ms
- Response opened
- 296 ms
- First visible word
- 724 ms
- Answer complete
- 814 ms
Montreal836 ms
Montreal- Finding the server
- 1 ms
- Reaching the server
- 2 ms
- Setting up security
- 3 ms
- Waiting for the server
- 393 ms
- Receiving the response: most of the time
- 438 ms
- Total
- 836 ms
- Response opened
- 339 ms
- First visible word
- 761 ms
- Answer complete
- 836 ms
Singapore504 ms
Singapore- Finding the server
- 3 ms
- Reaching the server
- 3 ms
- Setting up security
- 5 ms
- Waiting for the server
- 100 ms
- Receiving the response: most of the time
- 396 ms
- Total
- 504 ms
- Response opened
- 119 ms
- First visible word
- 436 ms
- Answer complete
- 504 ms
Tokyo708 ms
Tokyo- Finding the server
- 1 ms
- Reaching the server
- 2 ms
- Setting up security
- 14 ms
- Waiting for the server
- 191 ms
- Receiving the response: most of the time
- 501 ms
- Total
- 708 ms
- Response opened
- 196 ms
- First visible word
- 625 ms
- Answer complete
- 708 ms
Mumbai726 ms
Mumbai- Finding the server
- 8 ms
- Reaching the server
- 2 ms
- Setting up security
- 6 ms
- Waiting for the server
- 175 ms
- Receiving the response: most of the time
- 543 ms
- Total
- 726 ms
- Response opened
- 179 ms
- First visible word
- 662 ms
- Answer complete
- 726 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
DeepSeek still feels slow?
- Check DeepSeek’s status page
- Incidents and degraded performance are posted there. A slow first token during one is not your region.
- When every city slows down at once
- No single day in this window slowed every city: the slowest days were spread between 27 September and 2 October, and no city's daily median went above 0.91 seconds. When every city slows down on the same day, the cause is more likely on DeepSeek's side than on your route to it.
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 | 745 ms | 774 ms | 626 ms | 536 ms | 596 ms | 557 ms |
| 12 September 2026 | 714 ms | 631 ms | 791 ms | 658 ms | 395 ms | 613 ms |
| 13 September 2026 | 789 ms | 583 ms | 825 ms | 590 ms | 688 ms | 664 ms |
| 14 September 2026 | 823 ms | 837 ms | 923 ms | 701 ms | 780 ms | 737 ms |
| 15 September 2026 | 891 ms | 912 ms | 942 ms | 781 ms | 1406 ms ▲ | 867 ms |
| 16 September 2026 | 734 ms | 732 ms | 932 ms | 650 ms | 605 ms | 761 ms |
| 17 September 2026 | 670 ms | 810 ms | 662 ms | 492 ms | 419 ms | 525 ms |
| 18 September 2026 | 771 ms | 798 ms | 692 ms | 631 ms | 550 ms | 800 ms |
| 19 September 2026 | 670 ms | 690 ms | 726 ms | 459 ms | 526 ms | 589 ms |
| 20 September 2026 | 662 ms | 734 ms | 704 ms | 454 ms | 552 ms | 575 ms |
| 21 September 2026 | 735 ms | 876 ms | 802 ms | 665 ms | 569 ms | 656 ms |
| 22 September 2026 | 616 ms | 724 ms | 660 ms | 447 ms | 575 ms | 687 ms |
| 23 September 2026 | 780 ms | 655 ms | 679 ms | 530 ms | 558 ms | 710 ms |
| 24 September 2026 | 809 ms | 810 ms | 683 ms | 428 ms | 525 ms | 610 ms |
| 25 September 2026 | 629 ms | 568 ms | 628 ms | 340 ms | 547 ms | 457 ms |
| 26 September 2026 | 746 ms | 659 ms | 651 ms | 575 ms | 600 ms | 745 ms |
| 27 September 2026 | 821 ms | 672 ms | 734 ms | 467 ms | 629 ms | 725 ms |
| 28 September 2026 | 609 ms | 672 ms | 771 ms | 527 ms | 704 ms | 524 ms |
| 29 September 2026 | 685 ms | 575 ms | 863 ms | 711 ms | 638 ms | 539 ms |
| 30 September 2026 | 833 ms | 849 ms | 619 ms | 568 ms | 781 ms | 583 ms |
| 1 October 2026 | 813 ms | 710 ms | 711 ms | 460 ms | 583 ms | 565 ms |
| 2 October 2026 | 683 ms | 568 ms | 906 ms | 642 ms | 736 ms | 686 ms |
| 3 October 2026 | 738 ms | 811 ms | 697 ms | 511 ms | 597 ms | 724 ms |
| 4 October 2026 | 500 ms | 726 ms | 858 ms | 541 ms | 662 ms | 469 ms |
| 5 October 2026 | 604 ms | 811 ms | 699 ms | 597 ms | 480 ms | 692 ms |
| 6 October 2026 | 766 ms | 716 ms | 758 ms | 594 ms | 455 ms | 598 ms |
| 7 October 2026 | 700 ms | 750 ms | 754 ms | 521 ms | 607 ms | 537 ms |
| 9 October 2026 | 709 ms | 770 ms | 634 ms | 548 ms | 627 ms | 716 ms |
| 10 October 2026 | 591 ms | 667 ms | 664 ms | 409 ms | 403 ms | 524 ms |
About this measurement
POST api.deepseek.com/chat/completions · 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 deepseek-v4-flash on a one-word prompt: streaming, thinking disabled, temperature 0, 256 max tokens, paid key.
How LatencyRadar measures response time →
Technical details
Doesn’t measure: DeepSeek with thinking on, the chat app, tokens per second or long prompts. Not the same request as the low-effort direct providers.
One DeepSeek surface: time to first visible token from deepseek-v4-flash at api.deepseek.com/chat/completions, a one-word prompt, streaming, thinking disabled, temperature 0, 256 max tokens, paid key. Thinking is off because DeepSeek V4 Flash thinks at high effort by default, and the request is kept identical to the OpenRouter DeepSeek V4 Flash routes on the provider matrix, so you can read direct and routed side by side.
The ChatGPT, Claude and Grok figures on this page are the same prompt on the same days at low reasoning effort. Their daily medians ran 0.6 to 1.7 seconds (ChatGPT), 0.8 to 2.1 seconds (Claude) and 1.7 to 4.2 seconds (Grok) while DeepSeek's ran 0.4 to 0.9 seconds. Part of that gap is thinking this request turns off, so the two sets of numbers describe different requests and are not a ranking.
Typical first-token times over 27 September to 10 October 2026 are 527 ms in Singapore to 734 ms in Montreal, with Mumbai at 589 ms, Tokyo at 622 ms, Amsterdam at 700 ms and San Francisco at 728 ms. Every city shares one CloudFront address and the wait is shortest from Asia, which suggests that is where the model runs. Day to day, medians moved inside a band of 260 to 380 ms in every city. In the controlled run the slower responses sat 70 to 320 ms above that run's median, the widest gap in Singapore (753 ms against 436 ms).
| Test location | Typical response time | Slower response time | Requests |
|---|---|---|---|
| Amsterdam | 700 ms | 918 ms | 65 |
| San Francisco | 728 ms | 936 ms | 65 |
| Montreal | 734 ms | 990 ms | 65 |
| Singapore | 527 ms | 752 ms | 65 |
| Tokyo | 622 ms | 846 ms | 65 |
| Mumbai | 589 ms | 834 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.deepseek.com/chat/completions
- 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
- deepseek-v4-flash · thinking disabled · temperature 0 · max_tokens 256
- Prompt
- "Say 'ok' and nothing else." Identical across every city and every LLM API we benchmark
- Comparison figures
- ChatGPT (gpt-5.6-sol), Claude (claude-opus-5) and Grok (grok-4.6), all at low reasoning effort, from the same daily collection between 27 September and 10 October 2026. Those three think briefly before the first token; this request tells DeepSeek not to
- 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.
- 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.
DeepSeek API
https://api.deepseek.com/chat/completions
Independent measurement by LatencyRadar. Not affiliated with DeepSeek. · How we measure (v1.3)
How fast is your API compared to DeepSeek'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.
