Make your business agentic, while staying compliant.

• Foundry Azure, open sourceDefault

  Kimi K3

  Where a request is processed: Outside the EU

  Pros

  • Open-weights models at open-model prices, billed through your Azure subscription; the same weights can also run on your own machine.
  • No vendor account of ours: the deployment and its key are yours, with Azure governance and access control.

  Cons

  • Prompts are processed outside the EU: a third party serves this model, and Microsoft does not include that hosting in its EU commitments.
  • Paid per use, this model can be processed anywhere in the world; its host offers no regional option in the EU.

  Benchmark scores

  • GPQA Diamond93.5%
  • Terminal-Bench 2.188.3%
  • SWE-Bench Verifiedno public figure

  Source: model card, fetched 2026-09-28.

• Foundry Azure EU, open source

  DeepSeek V4 Flash

  Where a request is processed: In the EU (Microsoft Azure)

  Pros

  • Prompts never leave the EU: Microsoft processes them only within EU member countries.
  • Sold and supported by Microsoft under one Azure agreement.

  Cons

  • Fewer models than where processing may happen anywhere in the world, and weaker than the default on the published benchmarks.
  • New models arrive here later than where processing may happen anywhere in the world.

  Benchmark scores

  • GPQA Diamond88.1%
  • Terminal-Bench 2.182.7%
  • SWE-Bench Verified79.0%

  Source: model card, fetched 2026-09-28.

  Source: model card, fetched 2026-11-03.

• Frontier models in Foundry

  Claude Opus 5.5

  Where a request is processed: Outside the EU (Microsoft Azure)

  Pros

  • The highest-ranked models on Foundry's quality index and the Artificial Analysis Intelligence Index.
  • Sold through Azure under one agreement, billed on your Azure invoice.

  Cons

  • Closed weights: you cannot move the model to your own machines.
  • Prompts are processed outside the EU; for some of these models the model's vendor, not Microsoft, is the data processor.

  Benchmark scores

  • GPQA Diamondno public figure
  • Terminal-Bench 2.1no public figure
  • SWE-Bench Verifiedno public figure

• Self-hosted VM in your Azure

  DeepSeek V4.1 Flash (open weights)

  Where a request is processed: Your own machine

  Pros

  • Ahead of the default on 3 of the 4 published scores they share.
  • Full isolation: no prompt ever leaves machines you run, in the region you chose.
  • Any open-weights model whose licence allows it, including your own fine-tuned one.

  Cons

  • You patch, scale and monitor it; Spot prices are cheaper but the machine can be evicted.
  • The largest models need a multi-GPU machine; the price list shows no H100 or A100 machine in Norway East today, so Sweden Central or West Europe it is.

  Benchmark scores

  • GPQA Diamond90.9%
  • Terminal-Bench 2.190.6%
  • SWE-Bench Verifiedno public figure

  Source: model card, fetched 2026-11-03.

• Direct access

  Claude Opus 4.8

  Where a request is processed: Outside the EU (the vendor's own API)

  Pros

  • A direct connection to the vendor's own API, with no Azure deployment to set up.
  • Billed by the vendor per token, on your own account with it.

  Cons

  • Prompts are processed outside Azure and outside the EU, by the model vendor: you need a DPIA and a transfer safeguard.
  • Closed weights: you cannot move the model to your own machines.

  Benchmark scores

  • GPQA Diamond91.0%
  • Terminal-Bench 2.184.6%
  • SWE-Bench Verifiedno public figure

  Source: model card, fetched 2026-09-28.

GPQA Diamond

Graduate-level science questions

Terminal-Bench 2.1

Multi-step tasks in a real command-line terminal

SWE-Bench Verified

Fixing real, human-checked issues in open-source code

Scores in percent, as published on each figure's source page; higher is better. Tap a score to open that page. Where no public page gives a score, the card says so: we never estimate one.