Our take
Azure OpenAI Service is an enterprise-grade gateway to OpenAI models with strong compliance credentials and measured throughput. It offers EU-region hosting, SOC 2 attestation, and zero-retention options, but no downloadable or non-OpenAI models.
Use this for regulated workloads requiring SOC 2, zero-retention, and EU data residency. Pick it for enterprise procurement and support around OpenAI model access, or for high-throughput coding tasks where GPT-5-1-codex-mini delivers 113 tps. Skip it if you need open-weight models, if you want to compare non-OpenAI families in one catalogue, or if your budget cannot absorb the steep jump from nano tiers to mini and standard variants.
- Strongest compliance and data-control posture in the tracked set: SOC 2 attestation, zero-retention option, no training on prompts, and an EU-region endpoint — all four verified.
- Highest measured throughput for coding workloads: GPT-5-1-codex-mini at 113 tps, 2.45× the throughput of GPT-5-4-mini and 16.1× that of the legacy GPT-3.5-turbo variant.
- Lowest entry price for GPT-5-nano access among the tracked tiers.
- Broad OpenAI model catalogue — 55 tracked offers spanning nano, mini, and standard variants across GPT-4, GPT-4.1, GPT-4o, and GPT-5 families.
- Legacy model pricing is an outlier: the older GPT-3.5-turbo variant costs 20× the nano entry cost on input and 5× on output.
- Throughput varies dramatically within the same model: GPT-4o-mini ranges from 25 tps to 63.5 tps, GPT-4.1-nano from 38 tps to 67 tps, and GPT-4.1-mini from 46 tps to 105 tps — tier selection critically affects performance.
- No open-weight or non-OpenAI models in tracked offers: all 55 offers are OpenAI-branded, with no Mistral, Meta, Google, or other downloadable catalogue entries.
Privacy & data handling
- ✓
- Yes
- ✗
- No
- No answer on record
The colour says whether the answer favours you, not whether it is a yes — not training on your prompts earns a green cross, no zero-retention option earns a red one.
Prompt retention: none.