# Fastest AI API from South America (São Paulo)
Updated July 23, 2026. Independent, provider-neutral edge latency (TTFB) and uptime by provider, measured from South America (São Paulo), updated automatically.

As of July 23, 2026, measured from South America (São Paulo), the fastest AI inference API by edge latency (time-to-first-byte) is **openrouter** at 54 ms p50 (n=58).

## Edge latency (time-to-first-byte)
| # | Provider | p50 | p95 | Uptime | Samples |
| --- | --- | --- | --- | --- | --- |
| 1 | openrouter | 54 ms | 65 ms | 100% | 58 |
| 2 | baseten | 150 ms | 164 ms | 100% | 35 |
| 3 | google | 154 ms | 644 ms | 100% | 58 |
| 4 | cohere | 158 ms | 169 ms | 100% | 58 |
| 5 | cerebras | 184 ms | 228 ms | 100% | 58 |
| 6 | friendli | 194 ms | 342 ms | 100% | 35 |
| 7 | perplexity | 194 ms | 247 ms | 100% | 35 |
| 8 | replicate | 196 ms | 474 ms | 100% | 35 |
| 9 | anthropic | 211 ms | 603 ms | 100% | 58 |
| 10 | groq | 226 ms | 255 ms | 100% | 58 |
| 11 | meta-llama | 227 ms | 242 ms | 100% | 35 |
| 12 | inference-net | 233 ms | 942 ms | 100% | 35 |
| 13 | openai | 234 ms | 386 ms | 100% | 58 |
| 14 | novita | 237 ms | 273 ms | 100% | 35 |
| 15 | writer | 241 ms | 290 ms | 100% | 35 |
| 16 | reka | 241 ms | 662 ms | 100% | 35 |
| 17 | mistral | 246 ms | 290 ms | 100% | 58 |
| 18 | together | 261 ms | 715 ms | 100% | 58 |
| 19 | fireworks | 261 ms | 273 ms | 100% | 58 |
| 20 | xai | 265 ms | 992 ms | 100% | 58 |
| 21 | siliconflow | 274 ms | 280 ms | 100% | 35 |
| 22 | ai21 | 316 ms | 342 ms | 100% | 35 |
| 23 | targon | 323 ms | 329 ms | 100% | 35 |
| 24 | kimi | 341 ms | 369 ms | 100% | 40 |
| 25 | baichuan | 343 ms | 1242 ms | 100% | 35 |
| 26 | nscale | 407 ms | 409 ms | 100% | 35 |
| 27 | ernie | 410 ms | 419 ms | 100% | 35 |
| 28 | sambanova | 411 ms | 427 ms | 97% | 35 |
| 29 | deepseek | 426 ms | 470 ms | 100% | 58 |
| 30 | minimax | 452 ms | 491 ms | 100% | 40 |
| 31 | nebius | 461 ms | 478 ms | 100% | 35 |
| 32 | aleph-alpha | 479 ms | 519 ms | 100% | 35 |
| 33 | yi-01ai | 496 ms | 525 ms | 100% | 35 |
| 34 | deepinfra | 588 ms | 652 ms | 100% | 35 |
| 35 | sarvam | 601 ms | 628 ms | 100% | 35 |
| 36 | upstage | 606 ms | 646 ms | 100% | 35 |
| 37 | sensenova | 621 ms | 1469 ms | 100% | 35 |
| 38 | glm | 649 ms | 656 ms | 100% | 40 |
| 39 | iflytek | 656 ms | 1567 ms | 100% | 35 |
| 40 | stepfun | 666 ms | 1541 ms | 100% | 35 |
| 41 | doubao | 695 ms | 2457 ms | 100% | 35 |
| 42 | qwen | 716 ms | 861 ms | 100% | 40 |
| 43 | featherless | 778 ms | 3520 ms | 97% | 35 |
| 44 | hyperbolic | 786 ms | 857 ms | 100% | 35 |
| 45 | hunyuan | 1657 ms | 1672 ms | 100% | 35 |


Source: https://llmlatency.dev/region/sa-east