CL-SN55 941 —
GPT-61S 890 —
GPT-56T 861 —
GPT-6S 837 -1.2%
CL-OP55X 831 -1.4%
MUSE-SPK 804 —
QWEN-38X 795 —
GPT-6A 782 —
CL-OP5H 724 -1.6%
CL-OP5X 722 -1.6%
KIMI-K3X 718 -0.1%
CL-OP46H 707 -1.7%
CL-FAB5H 705 -7.4%
CL-OP55H 705 -1.7%
CL-OP47H 696 -1.7%
GEM-37FH 661 -9.7%
GEM-38FH 656 -4.8%
GPT-56S 578 -0.5%
CL-OP47 543 -0.5%
INKL 531 —
GEM-3P 494 -0.2%
CL-OP46 490 -0.2%
GEM-31P 485 -0.4%
CL-OP48 483 -0.2%
GPT-6L 436 —
CL-SN55 941 —
GPT-61S 890 —
GPT-56T 861 —
GPT-6S 837 -1.2%
CL-OP55X 831 -1.4%
MUSE-SPK 804 —
QWEN-38X 795 —
GPT-6A 782 —
CL-OP5H 724 -1.6%
CL-OP5X 722 -1.6%
KIMI-K3X 718 -0.1%
CL-OP46H 707 -1.7%
CL-FAB5H 705 -7.4%
CL-OP55H 705 -1.7%
CL-OP47H 696 -1.7%
GEM-37FH 661 -9.7%
GEM-38FH 656 -4.8%
GPT-56S 578 -0.5%
CL-OP47 543 -0.5%
INKL 531 —
GEM-3P 494 -0.2%
CL-OP46 490 -0.2%
GEM-31P 485 -0.4%
CL-OP48 483 -0.2%
GPT-6L 436 —
← Back to feed

OpenRouter US In-Region Routing: DeepSeek V4 Pro and Kimi K3 Now Processable Inside American Data Centers

OpenRouter’s US In-Region Routing is now live, joining the EU tier that launched earlier this year. Requests sent to us.openrouter.ai are decrypted and served only by providers running in the United States — the originating lab is not involved in request handling.

What It Resolves

Chinese open-weight models have faced a consistent procurement problem at US enterprises: even when the weights are legally accessible and technically useful, routing inference traffic through Chinese infrastructure is a non-starter for legal, compliance, and IT security teams. In-Region Routing addresses that directly.

Three Chinese-origin models are currently available on the US tier:

  • DeepSeek V4 Pro — served by Baseten and Fireworks from US data centers
  • Kimi K3 — Moonshot’s 2.8T open-weight model with 93.4% SWE-bench Verified
  • GLM 5.2 — also available on the EU tier via Mistral’s European infrastructure

When a request goes through us.openrouter.ai, it hits whichever US-based provider hosts the model. DeepSeek’s servers in China are not in the data path. For GLM 5.2, EU routing routes through Mistral.

Access and Pricing

In-Region Routing is gated to Business and Enterprise plans. Standard OpenRouter per-token pricing applies — there is no premium for in-region routing. The endpoint change is the only integration difference:

https://us.openrouter.ai/api/v1   # US in-region
https://eu.openrouter.ai/api/v1   # EU in-region

Model selection on the in-region endpoint is limited to models with confirmed in-region hosting. Not every model on the standard endpoint is available in-region — organizations need to verify the in-region model list before migrating existing integrations.

The Open-Weight Angle

OpenRouter’s own traffic data shows open-weight models have grown as a share of total tokens for both US and EU originating requests. The growth is partly driven by cost — DeepSeek V4 Pro and Kimi K3 carry significantly lower per-token prices than frontier proprietary models — and partly by the emergence of Chinese labs producing competitive models.

In-Region Routing makes that cost case actionable for a category of enterprises that previously had to self-host to achieve the same data residency guarantee. Self-hosting Kimi K3’s 2.8T parameter weights requires non-trivial GPU infrastructure; routing through a US-hosted Fireworks or Baseten endpoint removes that barrier.

NVIDIA’s Nemotron 3 Ultra and Thinking Machines’ Inkling are also included in the in-region model set — US-lab open-weight models that have gained commercial traction but until now lacked a managed residency path on the OpenRouter platform.

The Stripe Context

OpenRouter announced it was joining Stripe in August. The in-region routing launch is the first major product push under that arrangement. Stripe’s enterprise sales surface and compliance infrastructure are natural distribution channels for features that reduce friction for regulated-industry customers, which is the primary market for data residency routing.