DeepSeek API vs. OpenRouter latency, measured daily

DEEPSEEK LATENCY BENCHMARK · Sept 2026

We tested both from six cities. The faster option depends heavily on where you call from.

What the data shows #

1. Finding 1 ### Location matters more than the routeDirect was faster from Singapore and Mumbai. OpenRouter was often faster from Montreal.

• Singapore and Mumbai: direct 621–664 ms, routes 782–1,348 ms
• Montreal: routes 422–692 ms, direct 732 ms See the chart by city →

2. Finding 2 ### DeepSeek's direct API was more consistentDeepSeek's API varied less between locations: its slowest city was 36% above its fastest, against 89–219% through OpenRouter.

• City spread: 36% direct, 89–219% via OpenRouter
• Slow days: 0 direct, 1 via OpenRouter See the day-by-day chart →

3. Finding 3 ### Most of the wait isn't the networkConnection setup was only 11–19 ms. Most latency came after reaching the provider.

  - Network: 11–19 ms
  - First token: 462–883 ms
 [See where the time goes →](#time-split)

Time to first token, day by day #

A dot marks a day when every compared city was well above that provider’s usual level.

Show the numbers #

Does your location matter? #

DeepSeek direct was faster from Singapore and Mumbai in our tests.

• DeepSeek direct
• OpenRouter · Cloudflare
• OpenRouter · Together
• OpenRouter · DigitalOcean

Show exact values #

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.

Direct vs OpenRouter #

9–22 September 2026 · six test locations

Direct uses deepseek-v4-flash; the OpenRouter routes use deepseek-v4-flash-0731. Methodology →

Show full benchmark details ↓ #

| Detail | DeepSeek direct | OpenRouter · Cloudflare | OpenRouter · Together | OpenRouter · DigitalOcean |

|---|---|---|---|---|
| Model | deepseek-v4-flash | deepseek-v4-flash-0731 | deepseek-v4-flash-0731 | deepseek-v4-flash-0731 | 

| Slower requests, p95 (median city) | 973 ms | 1820 ms | 5649 ms | 2039 ms | | Fastest location | Tokyo · 575 ms | Amsterdam · 537 ms | San Francisco · 627 ms | Montreal · 422 ms | | Slowest location | San Francisco · 781 ms | Mumbai · 1211 ms | Mumbai · 1183 ms | Mumbai · 1348 ms | | Days when every city slowed together | 0 | 0 | 1 | 0 | | Requests in this window | 420 over 14 daily runs | 417 over 14 daily runs | 326 over 11 daily runs | 419 over 14 daily runs | | Latest controlled run | 7 September 2026 | 20 September 2026 | 13 September 2026 | 20 September 2026 |

How we test #

We send the same streaming prompt from six cities to DeepSeek directly and through OpenRouter, then measure time to first visible token.

Same request on both sides: thinking off, temperature 0. The Fireworks and BaseTen routes are left out: one throttled us, the other never returned a visible token.

Full testing methodology →

Good to know #

• Network setup is the hop to OpenRouter's edge (11–19 ms from every city); the hop from OpenRouter to the provider's machines is counted as waiting for the model, which is why Mumbai still waits longer than Montreal.
• Direct sends deepseek-v4-flash, which DeepSeek has served with V4.1 Flash since 10 September (V4 Flash is retired), while the OpenRouter routes pin deepseek-v4-flash-0731, so the two sides have not been the same model since that day.
• Each OpenRouter request pins one provider with fallbacks off, and a city's run is discarded if any response came from another provider.
• The prompt is "Say 'ok' and nothing else." (max_tokens 256, temperature 0, thinking off), sent daily between 03:30 and 06:46 UTC, 5 requests per city after one discarded warm-up.
• We time the first visible token only; tokens per second, long prompts and thinking on are not measured.

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.