Skip to main content
Watch: I Built an AI App Without Leaving Databricks

Cloud platforms - 68 entries

Azure OpenAI quotas and limits

Microsoft's official docs, read on October 17, 2026, set quotas and limits for Azure OpenAI in Foundry. For codex-mini, Tier 1 is listed at 1,000 RPM and 1,000,000 TPM (GlobalStandard), and Tier 6 at 15,000 RPM and 15,000,000 TPM (GlobalStandard).

Last verified from learn.microsoft.com. What the docs leave out

Sources
2
Tables
68
Entries
68
Verified
Oct 17, 2026

Limits by model, endpoint and plan

Values are copied exactly as the docs show them, units included, in the docs' own order. A dash means the docs list no value for that row.

All 68 tables on this page
  1. Azure OpenAI in Foundry Models (Global Standard, Data Zone Standard, Standard)
  2. codex-mini
  3. computer-use-preview
  4. gpt-4.1
  5. gpt-4.1-mini
  6. gpt-4.1-nano
  7. gpt-4o
  8. gpt-4o-audio-preview
  9. gpt-4o-mini
  10. gpt-4o-mini-audio-preview
  11. gpt-4o-mini-realtime-preview
  12. gpt-4o-realtime-preview
  13. gpt-5
  14. gpt-5-chat
  15. gpt-5-codex
  16. gpt-5-mini
  17. gpt-5-nano
  18. gpt-5-pro
  19. gpt-5.1
  20. gpt-5.1-chat
  21. gpt-5.1-codex
  22. gpt-5.1-codex-max
  23. gpt-5.1-codex-mini
  24. gpt-5.2
  25. gpt-5.2-chat
  26. gpt-5.3-chat
  27. gpt-5.2-codex
  28. gpt-5.3-codex
  29. gpt-5.4
  30. gpt-5.4-pro
  31. gpt-5.4-mini
  32. gpt-5.4-nano
  33. gpt-5.5
  34. gpt-5.6-luna
  35. gpt-5.6-sol
  36. gpt-5.6-terra
  37. gpt-6-astra
  38. gpt-6-luna
  39. gpt-6-sol
  40. gpt-6.1-sol
  41. gpt-chat-latest
  42. gpt-audio
  43. gpt-image-1
  44. gpt-image-1-mini
  45. gpt-image-1.5
  46. gpt-image-2
  47. gpt-image-2.5-flare
  48. gpt-image-2.5-sunburst
  49. gpt-realtime
  50. model-router
  51. o1
  52. o3
  53. o3-deep-research
  54. o3-mini
  55. o3-pro
  56. o4-mini
  57. text-embedding-3-large
  58. text-embedding-3-small
  59. GPT-Live API
  60. Azure OpenAI resource and API limits (quick reference)
  61. Azure OpenAI Batch API (input files)
  62. gpt-4-turbo
  63. gpt-4 (usage tier covers gpt-4 + gpt-4-32k, all versions)
  64. Usage tiers (Global Standard, Data Zone Standard, Standard)
  65. gpt-oss-120b
  66. Grok 4.7 version 1 (Foundry Models)
  67. Microsoft-Decision-1 (Foundry Models)
  68. Foundry Models sold by Azure not onboarded to subscription-level quota (per region, per subscription)

Azure OpenAI in Foundry Models (Global Standard, Data Zone Standard, Standard)

Quota tiers: Free Tier (Tier 0) and Tier 1-6

Azure OpenAI in Foundry Models (Global Standard, Data Zone Standard, Standard) limits
MetricQuota tiers: Free Tier (Tier 0) and Tier 1-6
Number of quota tiersSeven tiers are available: Free Tier and Tiers 1 through 6 - with Tier 6 offering the highest quotas
Quota scopeAzure subscription level (not tenant). Global Standard: one pool per model and version across all regions in a subscription. Data Zone Standard: one pool per model and version per data zone (for example, US or EU)
Subscription-level quota management startafter May 7, 2026
  • Replaces the former Default and Enterprise quota levels for pay as you go; previously approved quota increases are retained. Per-tier tables label the free tier 'Tier 0'. Per-model records below use the format 'Tier N <DeploymentType> RPM | TPM' with cells copied verbatim; '/ 10s' means the published RPM cell is per 10 seconds; '-' means no TPM published. Batch metric order: Enterprise and MCA-E / Default / Monthly credit card-based subscriptions / MSDN subscriptions / Azure for Students, free trials (B = billion, M = million, K = thousand). Per-region availability is shown in the Foundry portal (https://ai.azure.com/resource/quota) and via the Usages API and Model Capacities API. Old docs URL learn.microsoft.com/azure/ai-foundry/openai/quotas-limits now serves /azure/foundry/openai/quotas-limits (ms.date 2026-08-20).

codex-mini

6 tiers - Quota Tier 0-6

codex-mini limits
TierGlobalStandard RPM | TPM
Tier 11,000 | 1,000,000
Tier 22,000 | 2,000,000
Tier 34,000 | 4,000,000
Tier 47,000 | 7,000,000
Tier 510,000 | 10,000,000
Tier 615,000 | 15,000,000

computer-use-preview

6 tiers - Quota Tier 0-6

computer-use-preview limits
TierGlobalStandard RPM | TPM
Tier 14,500 | 450,000
Tier 220,000 | 2,000,000
Tier 370,000 | 7,000,000
Tier 4160,000 | 16,000,000
Tier 5300,000 | 30,000,000
Tier 6450,000 | 45,000,000

gpt-4.1

6 tiers - Quota Tier 0-6Model page: GPT-4.1

gpt-4.1 limits
TierDataZoneStandard RPM | TPM
Tier 1300 | 300,000
GlobalStandard RPM | TPM
1,000 | 1,000,000
Tier 21,000 | 1,000,000
GlobalStandard RPM | TPM
3,000 | 3,000,000
Tier 33,000 | 3,000,000
GlobalStandard RPM | TPM
9,000 | 9,000,000
Tier 46,000 | 6,000,000
GlobalStandard RPM | TPM
18,000 | 18,000,000
Tier 510,000 | 10,000,000
GlobalStandard RPM | TPM
30,000 | 30,000,000
Tier 615,000 | 15,000,000
GlobalStandard RPM | TPM
45,000 | 45,000,000
gpt-4.1, other limits
MetricLimit
Usage tier per month (latency threshold)30 billion tokens
Global batch enqueued tokens (Enterprise and MCA-E / Default / Credit card / MSDN / Students, trials)5B / 200M / 50M / 90K / N/A
Data zone batch enqueued tokens (Enterprise and MCA-E / Default / Credit card / MSDN / Students, trials)500M / 30M / 30M / 90K / N/A

gpt-4.1-mini

7 tiers - Quota Tier 0-6Model page: GPT-4.1 Mini

gpt-4.1-mini limits
TierGlobalStandard RPM | TPM
Tier 0200 | 200,000
Tier 15,000 | 5,000,000
DataZoneStandard RPM | TPM
2,000 | 2,000,000
Standard RPM | TPM
6,000 | 6,000,000
Tier 216,000 | 16,000,000
DataZoneStandard RPM | TPM
6,000 | 6,000,000
Standard RPM | TPM
12,000 | 12,000,000
Tier 346,000 | 46,000,000
DataZoneStandard RPM | TPM
16,000 | 16,000,000
Standard RPM | TPM
30,000 | 30,000,000
Tier 490,000 | 90,000,000
DataZoneStandard RPM | TPM
31,000 | 31,000,000
Standard RPM | TPM
75,000 | 75,000,000
Tier 5150,000 | 150,000,000
DataZoneStandard RPM | TPM
50,000 | 50,000,000
Standard RPM | TPM
150,000 | 150,000,000
Tier 6225,000 | 225,000,000
DataZoneStandard RPM | TPM
75,000 | 75,000,000
Standard RPM | TPM
225,000 | 225,000,000
gpt-4.1-mini, other limits
MetricLimit
Usage tier per month (latency threshold)150 billion tokens
Global batch enqueued tokens (Enterprise and MCA-E / Default / Credit card / MSDN / Students, trials)15B / 1B / 50M / 90K / N/A
Data zone batch enqueued tokens (Enterprise and MCA-E / Default / Credit card / MSDN / Students, trials)1.5B / 100M / 50M / 90K / N/A
  • Global batch table writes the model as `gpt-4.1 mini`. One of four models with a published Tier 0 (Free Tier) row.

gpt-4.1-nano

6 tiers - Quota Tier 0-6Model page: GPT-4.1 nano

gpt-4.1-nano limits
TierDataZoneStandard RPM | TPM
Tier 12,000 | 2,000,000
GlobalStandard RPM | TPM
5,000 | 5,000,000
Tier 26,000 | 6,000,000
GlobalStandard RPM | TPM
16,000 | 16,000,000
Tier 316,000 | 16,000,000
GlobalStandard RPM | TPM
46,000 | 46,000,000
Tier 431,000 | 31,000,000
GlobalStandard RPM | TPM
90,000 | 90,000,000
Tier 550,000 | 50,000,000
GlobalStandard RPM | TPM
150,000 | 150,000,000
Tier 675,000 | 75,000,000
GlobalStandard RPM | TPM
225,000 | 225,000,000
gpt-4.1-nano, other limits
MetricLimit
Usage tier per month (latency threshold)550 billion tokens
Global batch enqueued tokens (Enterprise and MCA-E / Default / Credit card / MSDN / Students, trials)15B / 1B / 50M / 90K / N/A

gpt-4o

6 tiers - Quota Tier 0-6Model page: GPT-4o

gpt-4o limits
TierDataZoneStandard RPM | TPM
Tier 1300 / 10s | 300,000
Tier 21000 / 10s | 1,000,000
Tier 33000 / 10s | 3,000,000
Tier 46000 / 10s | 6,000,000
Tier 510000 / 10s | 10,000,000
Tier 615000 / 10s | 15,000,000
gpt-4o, other limits
MetricLimit
Usage tier per month (latency threshold)12 billion tokens
Global batch enqueued tokens (Enterprise and MCA-E / Default / Credit card / MSDN / Students, trials)5B / 200M / 50M / 90K / N/A
Data zone batch enqueued tokens (Enterprise and MCA-E / Default / Credit card / MSDN / Students, trials)500M / 30M / 30M / 90K / N/A
  • Only DataZoneStandard rows are published for gpt-4o in the tier tables.

gpt-4o-audio-preview

6 tiers - Quota Tier 0-6

gpt-4o-audio-preview limits
TierGlobalStandard RPM | TPM
Tier 130000 / 10s | 30,000,000
Tier 230000 / 10s | 30,000,000
Tier 330000 / 10s | 30,000,000
Tier 430000 / 10s | 30,000,000
Tier 530000 / 10s | 30,000,000
Tier 645000 / 10s | 45,000,000

gpt-4o-mini

6 tiers - Quota Tier 0-6Model page: GPT-4o Mini

gpt-4o-mini limits
TierDataZoneStandard RPM | TPM
Tier 110,000 | 1,000,000
GlobalStandard RPM | TPM
20,000 | 2,000,000
Tier 230,000 | 3,000,000
GlobalStandard RPM | TPM
90,000 | 9,000,000
Tier 370,000 | 7,000,000
GlobalStandard RPM | TPM
330,000 | 33,000,000
Tier 4130,000 | 13,000,000
GlobalStandard RPM | TPM
780,000 | 78,000,000
Tier 5200,000 | 20,000,000
GlobalStandard RPM | TPM
1,500,000 | 150,000,000
Tier 6300,000 | 30,000,000
GlobalStandard RPM | TPM
2,250,000 | 225,000,000
gpt-4o-mini, other limits
MetricLimit
Usage tier per month (latency threshold)85 billion tokens
Global batch enqueued tokens (Enterprise and MCA-E / Default / Credit card / MSDN / Students, trials)15B / 1B / 50M / 90K / N/A
Data zone batch enqueued tokens (Enterprise and MCA-E / Default / Credit card / MSDN / Students, trials)1.5B / 100M / 50M / 90K / N/A

gpt-4o-mini-audio-preview

6 tiers - Quota Tier 0-6

gpt-4o-mini-audio-preview limits
TierGlobalStandard RPM | TPM
Tier 130000 / 10s | 30,000,000
Tier 230000 / 10s | 30,000,000
Tier 330000 / 10s | 30,000,000
Tier 430000 / 10s | 30,000,000
Tier 530000 / 10s | 30,000,000
Tier 645000 / 10s | 45,000,000

gpt-4o-mini-realtime-preview

6 tiers - Quota Tier 0-6

gpt-4o-mini-realtime-preview limits
TierGlobalStandard RPM | TPM
Tier 136 | 6,000
Tier 236 | 6,000
Tier 336 | 6,000
Tier 436 | 6,000
Tier 536 | 6,000
Tier 654 | 9,000

gpt-4o-realtime-preview

6 tiers - Quota Tier 0-6

gpt-4o-realtime-preview limits
TierGlobalStandard RPM | TPM
Tier 136 | 6,000
Tier 236 | 6,000
Tier 336 | 6,000
Tier 436 | 6,000
Tier 536 | 6,000
Tier 654 | 9,000

gpt-5

6 tiers - Quota Tier 0-6Model page: GPT-5

gpt-5 limits
TierDataZoneStandard RPM | TPM
Tier 13,000 | 300,000
GlobalStandard RPM | TPM
10,000 | 1,000,000
Tier 210,000 | 1,000,000
GlobalStandard RPM | TPM
30,000 | 3,000,000
Tier 330,000 | 3,000,000
GlobalStandard RPM | TPM
90,000 | 9,000,000
Tier 460,000 | 6,000,000
GlobalStandard RPM | TPM
180,000 | 18,000,000
Tier 5100,000 | 10,000,000
GlobalStandard RPM | TPM
300,000 | 30,000,000
Tier 6150,000 | 15,000,000
GlobalStandard RPM | TPM
450,000 | 45,000,000
gpt-5, other limits
MetricLimit
Usage tier per month (latency threshold)32 billion tokens
Global batch enqueued tokens (Enterprise and MCA-E / Default / Credit card / MSDN / Students, trials)5B / 200M / 50M / 90K / N/A
Data zone batch enqueued tokens (Enterprise and MCA-E / Default / Credit card / MSDN / Students, trials)5B / 200M / 50M / 90K / N/A

gpt-5-chat

6 tiers - Quota Tier 0-6

gpt-5-chat limits
TierGlobalStandard RPM | TPM
Tier 11,000 | 1,000,000
Tier 22,000 | 2,000,000
Tier 33,000 | 3,000,000
Tier 44,000 | 4,000,000
Tier 55,000 | 5,000,000
Tier 68,000 | 8,000,000
gpt-5-chat, other limits
MetricLimit
Usage tier per month (latency threshold)32 billion tokens

gpt-5-codex

6 tiers - Quota Tier 0-6

gpt-5-codex limits
TierGlobalStandard RPM | TPM
Tier 11,000 | 1,000,000
Tier 22,000 | 2,000,000
Tier 34,000 | 4,000,000
Tier 47,000 | 7,000,000
Tier 510,000 | 10,000,000
Tier 615,000 | 15,000,000

gpt-5-mini

7 tiers - Quota Tier 0-6Model page: GPT-5 Mini

gpt-5-mini limits
TierGlobalStandard RPM | TPM
Tier 0500 | 500,000
Tier 11,000 | 1,000,000
DataZoneStandard RPM | TPM
300 | 300,000
Tier 22,000 | 2,000,000
DataZoneStandard RPM | TPM
670 | 670,000
Tier 34,000 | 4,000,000
DataZoneStandard RPM | TPM
1,000 | 1,000,000
Tier 47,000 | 7,000,000
DataZoneStandard RPM | TPM
2,000 | 2,000,000
Tier 510,000 | 10,000,000
DataZoneStandard RPM | TPM
3,000 | 3,000,000
Tier 615,000 | 15,000,000
DataZoneStandard RPM | TPM
4,000 | 4,000,000
gpt-5-mini, other limits
MetricLimit
Usage tier per month (latency threshold)160 billion tokens
  • One of four models with a published Tier 0 (Free Tier) row.

gpt-5-nano

6 tiers - Quota Tier 0-6Model page: GPT-5 nano

gpt-5-nano limits
TierDataZoneStandard RPM | TPM
Tier 12,000 | 2,000,000
GlobalStandard RPM | TPM
5,000 | 5,000,000
Tier 26,000 | 6,000,000
GlobalStandard RPM | TPM
16,000 | 16,000,000
Tier 316,000 | 16,000,000
GlobalStandard RPM | TPM
46,000 | 46,000,000
Tier 431,000 | 31,000,000
GlobalStandard RPM | TPM
90,000 | 90,000,000
Tier 550,000 | 50,000,000
GlobalStandard RPM | TPM
150,000 | 150,000,000
Tier 675,000 | 75,000,000
GlobalStandard RPM | TPM
225,000 | 225,000,000
gpt-5-nano, other limits
MetricLimit
Usage tier per month (latency threshold)800 billion tokens

gpt-5-pro

6 tiers - Quota Tier 0-6Model page: GPT-5 Pro

gpt-5-pro limits
TierGlobalStandard RPM | TPM
Tier 11,600 | 160,000
Tier 23,500 | 350,000
Tier 37,000 | 700,000
Tier 411,500 | 1,150,000
Tier 516,000 | 1,600,000
Tier 624,000 | 2,400,000

gpt-5.1

6 tiers - Quota Tier 0-6Model page: GPT-5.1

gpt-5.1 limits
TierDataZoneStandard RPM | TPM
Tier 13,000 | 300,000
GlobalStandard RPM | TPM
10,000 | 1,000,000
Standard RPM | TPM
3,000 | 300,000
Tier 26,700 | 670,000
GlobalStandard RPM | TPM
20,000 | 2,000,000
Standard RPM | TPM
6,700 | 670,000
Tier 310,000 | 1,000,000
GlobalStandard RPM | TPM
40,000 | 4,000,000
Standard RPM | TPM
10,000 | 1,000,000
Tier 420,000 | 2,000,000
GlobalStandard RPM | TPM
70,000 | 7,000,000
Standard RPM | TPM
20,000 | 2,000,000
Tier 530,000 | 3,000,000
GlobalStandard RPM | TPM
100,000 | 10,000,000
Standard RPM | TPM
30,000 | 3,000,000
Tier 640,000 | 4,000,000
GlobalStandard RPM | TPM
150,000 | 15,000,000
Standard RPM | TPM
40,000 | 4,000,000
gpt-5.1, other limits
MetricLimit
Usage tier per month (latency threshold)86 billion tokens
Global batch enqueued tokens (Enterprise and MCA-E / Default / Credit card / MSDN / Students, trials)5B / 200M / 50M / 90K / N/A
Data zone batch enqueued tokens (Enterprise and MCA-E / Default / Credit card / MSDN / Students, trials)5B / 200M / 50M / 90K / N/A

gpt-5.1-chat

6 tiers - Quota Tier 0-6

gpt-5.1-chat limits
TierGlobalStandard RPM | TPM
Tier 110,000 | 1,000,000
Tier 220,000 | 2,000,000
Tier 330,000 | 3,000,000
Tier 440,000 | 4,000,000
Tier 550,000 | 5,000,000
Tier 680,000 | 8,000,000

gpt-5.1-codex

6 tiers - Quota Tier 0-6

gpt-5.1-codex limits
TierDataZoneStandard RPM | TPM
Tier 13,000 | 300,000
GlobalStandard RPM | TPM
1,000 | 1,000,000
Tier 26,700 | 670,000
GlobalStandard RPM | TPM
2,000 | 2,000,000
Tier 310,000 | 1,000,000
GlobalStandard RPM | TPM
4,000 | 4,000,000
Tier 420,000 | 2,000,000
GlobalStandard RPM | TPM
7,000 | 7,000,000
Tier 530,000 | 3,000,000
GlobalStandard RPM | TPM
10,000 | 10,000,000
Tier 640,000 | 4,000,000
GlobalStandard RPM | TPM
15,000 | 15,000,000
gpt-5.1-codex, other limits
MetricLimit
Usage tier per month (latency threshold)86 billion tokens

gpt-5.1-codex-max

6 tiers - Quota Tier 0-6

gpt-5.1-codex-max limits
TierGlobalStandard RPM | TPM
Tier 110,000 | 1,000,000
Tier 220,000 | 2,000,000
Tier 340,000 | 4,000,000
Tier 470,000 | 7,000,000
Tier 5100,000 | 10,000,000
Tier 6150,000 | 15,000,000

gpt-5.1-codex-mini

6 tiers - Quota Tier 0-6

gpt-5.1-codex-mini limits
TierGlobalStandard RPM | TPM
Tier 11,000 | 1,000,000
Tier 22,000 | 2,000,000
Tier 34,000 | 4,000,000
Tier 47,000 | 7,000,000
Tier 510,000 | 10,000,000
Tier 615,000 | 15,000,000

gpt-5.2

6 tiers - Quota Tier 0-6Model page: GPT-5.2

gpt-5.2 limits
TierDataZoneStandard RPM | TPM
Tier 13,000 | 300,000
GlobalStandard RPM | TPM
10,000 | 1,000,000
Tier 26,700 | 670,000
GlobalStandard RPM | TPM
20,000 | 2,000,000
Tier 310,000 | 1,000,000
GlobalStandard RPM | TPM
40,000 | 4,000,000
Tier 420,000 | 2,000,000
GlobalStandard RPM | TPM
70,000 | 7,000,000
Tier 530,000 | 3,000,000
GlobalStandard RPM | TPM
100,000 | 10,000,000
Tier 640,000 | 4,000,000
GlobalStandard RPM | TPM
150,000 | 15,000,000
gpt-5.2, other limits
MetricLimit
Usage tier per month (latency threshold)60 billion tokens
Global batch enqueued tokens (Enterprise and MCA-E / Default / Credit card / MSDN / Students, trials)5B / 200M / 50M / N/A / N/A

gpt-5.2-chat

6 tiers - Quota Tier 0-6

gpt-5.2-chat limits
TierGlobalStandard RPM | TPM
Tier 110,000 | 1,000,000
Tier 220,000 | 2,000,000
Tier 330,000 | 3,000,000
Tier 440,000 | 4,000,000
Tier 550,000 | 5,000,000
Tier 680,000 | 8,000,000

gpt-5.3-chat

6 tiers - Quota Tier 0-6

gpt-5.3-chat limits
TierGlobalStandard RPM | TPM
Tier 11,000 | 1,000,000
Tier 22,000 | 2,000,000
Tier 33,000 | 3,000,000
Tier 44,000 | 4,000,000
Tier 55,000 | 5,000,000
Tier 68,000 | 8,000,000

gpt-5.2-codex

6 tiers - Quota Tier 0-6

gpt-5.2-codex limits
TierGlobalStandard RPM | TPM
Tier 110,000 | 1,000,000
Tier 220,000 | 2,000,000
Tier 340,000 | 4,000,000
Tier 470,000 | 7,000,000
Tier 5100,000 | 10,000,000
Tier 6150,000 | 15,000,000

gpt-5.3-codex

6 tiers - Quota Tier 0-6Model page: GPT-5.3-Codex

gpt-5.3-codex limits
TierGlobalStandard RPM | TPM
Tier 110,000 | 1,000,000
Tier 220,000 | 2,000,000
Tier 340,000 | 4,000,000
Tier 470,000 | 7,000,000
Tier 5100,000 | 10,000,000
Tier 6150,000 | 15,000,000
gpt-5.3-codex, other limits
MetricLimit
Usage tier per month (latency threshold)60 billion tokens

gpt-5.4

6 tiers - Quota Tier 0-6Model page: GPT-5.4

gpt-5.4 limits
TierDataZoneStandard RPM | TPM
Tier 1300 | 300,000
GlobalStandard RPM | TPM
10,000 | 1,000,000
Tier 2670 | 670,000
GlobalStandard RPM | TPM
20,000 | 2,000,000
Tier 31,000 | 1,000,000
GlobalStandard RPM | TPM
40,000 | 4,000,000
Tier 42,000 | 2,000,000
GlobalStandard RPM | TPM
70,000 | 7,000,000
Tier 53,000 | 3,000,000
GlobalStandard RPM | TPM
100,000 | 10,000,000
Tier 64,000 | 4,000,000
GlobalStandard RPM | TPM
150,000 | 15,000,000
gpt-5.4, other limits
MetricLimit
Usage tier per month (latency threshold)50 billion tokens
Global batch enqueued tokens (Enterprise and MCA-E / Default / Credit card / MSDN / Students, trials)5B / 200M / 50M / N/A / N/A
Data zone batch enqueued tokens (Enterprise and MCA-E / Default / Credit card / MSDN / Students, trials)5B / 200M / 50M / N/A / N/A

gpt-5.4-pro

6 tiers - Quota Tier 0-6

gpt-5.4-pro limits
TierGlobalStandard RPM | TPM
Tier 1160 | 160,000
Tier 2350 | 350,000
Tier 3700 | 700,000
Tier 41,150 | 1,150,000
Tier 51,600 | 1,600,000
Tier 62,400 | 2,400,000

gpt-5.4-mini

6 tiers - Quota Tier 0-6Model page: GPT-5.4 Mini

gpt-5.4-mini limits
TierGlobalStandard RPM | TPM
Tier 11,000 | 1,000,000
Tier 22,000 | 2,000,000
Tier 34,000 | 4,000,000
Tier 47,000 | 7,000,000
Tier 510,000 | 10,000,000
Tier 615,000 | 15,000,000
gpt-5.4-mini, other limits
MetricLimit
Usage tier per month (latency threshold)165 billion tokens
Global batch enqueued tokens (Enterprise and MCA-E / Default / Credit card / MSDN / Students, trials)5B / 200M / 50M / N/A / N/A
Data zone batch enqueued tokens (Enterprise and MCA-E / Default / Credit card / MSDN / Students, trials)5B / 200M / 50M / N/A / N/A

gpt-5.4-nano

6 tiers - Quota Tier 0-6Model page: GPT-5.4 nano

gpt-5.4-nano limits
TierDataZoneStandard RPM | TPM
Tier 12,000 | 2,000,000
GlobalStandard RPM | TPM
5,000 | 5,000,000
Tier 26,000 | 6,000,000
GlobalStandard RPM | TPM
16,000 | 16,000,000
Tier 316,000 | 16,000,000
GlobalStandard RPM | TPM
46,000 | 46,000,000
Tier 431,000 | 31,000,000
GlobalStandard RPM | TPM
90,000 | 90,000,000
Tier 550,000 | 50,000,000
GlobalStandard RPM | TPM
150,000 | 150,000,000
Tier 675,000 | 75,000,000
GlobalStandard RPM | TPM
225,000 | 225,000,000
gpt-5.4-nano, other limits
MetricLimit
Usage tier per month (latency threshold)605 billion tokens
Global batch enqueued tokens (Enterprise and MCA-E / Default / Credit card / MSDN / Students, trials)5B / 200M / 50M / N/A / N/A

gpt-5.5

6 tiers - Quota Tier 0-6Model page: GPT-5.5

gpt-5.5 limits
TierDataZoneStandard RPM | TPM
Tier 1333 | 333,000
GlobalStandard RPM | TPM
1,000 | 1,000,000
Tier 2667 | 667,000
GlobalStandard RPM | TPM
2,000 | 2,000,000
Tier 31,333 | 1,333,000
GlobalStandard RPM | TPM
4,000 | 4,000,000
Tier 42,333 | 2,333,000
GlobalStandard RPM | TPM
7,000 | 7,000,000
Tier 53,000 | 3,000,000
GlobalStandard RPM | TPM
10,000 | 10,000,000
Tier 65,000 | 5,000,000
GlobalStandard RPM | TPM
15,000 | 15,000,000
gpt-5.5, other limits
MetricLimit
Usage tier per month (latency threshold)25 billion tokens
Global batch enqueued tokens (Enterprise and MCA-E / Default / Credit card / MSDN / Students, trials)5B / 200M / 50M / 90K / N/A
Data zone batch enqueued tokens (Enterprise and MCA-E / Default / Credit card / MSDN / Students, trials)5B / 200M / 50M / 90K / N/A
  • Tier 5 DataZoneStandard is published as 3,000 | 3,000,000 for gpt-5.5 (other 333-base models show 3,333 | 3,333,000).

gpt-5.6-luna

6 tiers - Quota Tier 0-6Model page: GPT-5.6 Luna

gpt-5.6-luna limits
TierDataZoneStandard RPM | TPM
Tier 1333 | 333,000
GlobalStandard RPM | TPM
1,000 | 1,000,000
Tier 2667 | 667,000
GlobalStandard RPM | TPM
2,000 | 2,000,000
Tier 31,333 | 1,333,000
GlobalStandard RPM | TPM
4,000 | 4,000,000
Tier 42,333 | 2,333,000
GlobalStandard RPM | TPM
7,000 | 7,000,000
Tier 53,333 | 3,333,000
GlobalStandard RPM | TPM
10,000 | 10,000,000
Tier 65,000 | 5,000,000
GlobalStandard RPM | TPM
15,000 | 15,000,000

gpt-5.6-sol

6 tiers - Quota Tier 0-6Model page: GPT-5.6 Sol

gpt-5.6-sol limits
TierDataZoneStandard RPM | TPM
Tier 1333 | 333,000
GlobalStandard RPM | TPM
1,000 | 1,000,000
Tier 2667 | 667,000
GlobalStandard RPM | TPM
2,000 | 2,000,000
Tier 31,333 | 1,333,000
GlobalStandard RPM | TPM
4,000 | 4,000,000
Tier 42,333 | 2,333,000
GlobalStandard RPM | TPM
7,000 | 7,000,000
Tier 53,333 | 3,333,000
GlobalStandard RPM | TPM
10,000 | 10,000,000
Tier 65,000 | 5,000,000
GlobalStandard RPM | TPM
15,000 | 15,000,000
gpt-5.6-sol, other limits
MetricLimit
Usage tier per month (latency threshold)25 billion tokens

gpt-5.6-terra

6 tiers - Quota Tier 0-6Model page: GPT-5.6 Terra

gpt-5.6-terra limits
TierDataZoneStandard RPM | TPM
Tier 1333 | 333,000
GlobalStandard RPM | TPM
1,000 | 1,000,000
Tier 2667 | 667,000
GlobalStandard RPM | TPM
2,000 | 2,000,000
Tier 31,333 | 1,333,000
GlobalStandard RPM | TPM
4,000 | 4,000,000
Tier 42,333 | 2,333,000
GlobalStandard RPM | TPM
7,000 | 7,000,000
Tier 53,333 | 3,333,000
GlobalStandard RPM | TPM
10,000 | 10,000,000
Tier 65,000 | 5,000,000
GlobalStandard RPM | TPM
15,000 | 15,000,000

gpt-6-astra

6 tiers - Quota Tier 0-6Model page: GPT-6 Astra

gpt-6-astra limits
TierDataZoneStandard RPM | TPM
Tier 1333 | 333,000
GlobalStandard RPM | TPM
1,000 | 1,000,000
Tier 2667 | 667,000
GlobalStandard RPM | TPM
2,000 | 2,000,000
Tier 31,333 | 1,333,000
GlobalStandard RPM | TPM
4,000 | 4,000,000
Tier 42,333 | 2,333,000
GlobalStandard RPM | TPM
7,000 | 7,000,000
Tier 53,333 | 3,333,000
GlobalStandard RPM | TPM
10,000 | 10,000,000
Tier 65,000 | 5,000,000
GlobalStandard RPM | TPM
15,000 | 15,000,000

gpt-6-luna

6 tiers - Quota Tier 0-6Model page: GPT-6 Luna

gpt-6-luna limits
TierDataZoneStandard RPM | TPM
Tier 1333 | 333,000
GlobalStandard RPM | TPM
1,000 | 1,000,000
Tier 2667 | 667,000
GlobalStandard RPM | TPM
2,000 | 2,000,000
Tier 31,333 | 1,333,000
GlobalStandard RPM | TPM
4,000 | 4,000,000
Tier 42,333 | 2,333,000
GlobalStandard RPM | TPM
7,000 | 7,000,000
Tier 53,333 | 3,333,000
GlobalStandard RPM | TPM
10,000 | 10,000,000
Tier 65,000 | 5,000,000
GlobalStandard RPM | TPM
15,000 | 15,000,000

gpt-6-sol

6 tiers - Quota Tier 0-6Model page: GPT-6 Sol

gpt-6-sol limits
TierDataZoneStandard RPM | TPM
Tier 1333 | 333,000
GlobalStandard RPM | TPM
1,000 | 1,000,000
Tier 2667 | 667,000
GlobalStandard RPM | TPM
2,000 | 2,000,000
Tier 31,333 | 1,333,000
GlobalStandard RPM | TPM
4,000 | 4,000,000
Tier 42,333 | 2,333,000
GlobalStandard RPM | TPM
7,000 | 7,000,000
Tier 53,333 | 3,333,000
GlobalStandard RPM | TPM
10,000 | 10,000,000
Tier 65,000 | 5,000,000
GlobalStandard RPM | TPM
15,000 | 15,000,000

gpt-6.1-sol

6 tiers - Quota Tier 0-6Model page: GPT-6.1 Sol

gpt-6.1-sol limits
TierDataZoneStandard RPM | TPM
Tier 1333 | 333,000
GlobalStandard RPM | TPM
1,000 | 1,000,000
Tier 2667 | 667,000
GlobalStandard RPM | TPM
2,000 | 2,000,000
Tier 31,333 | 1,333,000
GlobalStandard RPM | TPM
4,000 | 4,000,000
Tier 42,333 | 2,333,000
GlobalStandard RPM | TPM
7,000 | 7,000,000
Tier 53,333 | 3,333,000
GlobalStandard RPM | TPM
10,000 | 10,000,000
Tier 65,000 | 5,000,000
GlobalStandard RPM | TPM
15,000 | 15,000,000

gpt-chat-latest

6 tiers - Quota Tier 0-6

gpt-chat-latest limits
TierGlobalStandard (versions 2026-05-05, 2026-05-28, 2026-06-24) RPM | TPM
Tier 110,000 | 1,000,000
GlobalStandard (versions 2026-08-06, 2026-10-07) RPM | TPM
1,000 | 1,000,000
DataZoneStandard (version 2026-10-07) RPM | TPM
333 | 333,000
Tier 220,000 | 2,000,000
GlobalStandard (versions 2026-08-06, 2026-10-07) RPM | TPM
2,000 | 2,000,000
DataZoneStandard (version 2026-10-07) RPM | TPM
667 | 667,000
Tier 330,000 | 3,000,000
GlobalStandard (versions 2026-08-06, 2026-10-07) RPM | TPM
3,000 | 3,000,000
DataZoneStandard (version 2026-10-07) RPM | TPM
1,000 | 1,000,000
Tier 440,000 | 4,000,000
GlobalStandard (versions 2026-08-06, 2026-10-07) RPM | TPM
4,000 | 4,000,000
DataZoneStandard (version 2026-10-07) RPM | TPM
1,333 | 1,333,000
Tier 550,000 | 5,000,000
GlobalStandard (versions 2026-08-06, 2026-10-07) RPM | TPM
5,000 | 5,000,000
DataZoneStandard (version 2026-10-07) RPM | TPM
1,667 | 1,667,000
Tier 680,000 | 8,000,000
GlobalStandard (versions 2026-08-06, 2026-10-07) RPM | TPM
8,000 | 8,000,000
DataZoneStandard (version 2026-10-07) RPM | TPM
2,667 | 2,667,000
  • For Global Standard, all gpt-chat-latest versions share the same tier-level TPM limit. Versions 2026-05-05, 2026-05-28 and 2026-06-24 use 10 RPM per 1,000 TPM; versions 2026-08-06 and 2026-10-07 use 1 RPM per 1,000 TPM. Version 2026-10-07 Data Zone Standard also uses 1 RPM per 1,000 TPM. Tables list these as gpt-chat-latest with footnotes 1, 2, 3.

gpt-audio

6 tiers - Quota Tier 0-6

gpt-audio limits
TierGlobalStandard RPM | TPM
Tier 130000 / 10s | 30,000,000
Tier 230000 / 10s | 30,000,000
Tier 330000 / 10s | 30,000,000
Tier 430000 / 10s | 30,000,000
Tier 530000 / 10s | 30,000,000
Tier 645000 / 10s | 45,000,000

gpt-image-1

6 tiers - Quota Tier 0-6

gpt-image-1 limits
TierGlobalStandard RPM | TPM
Tier 19 | -
Tier 218 | -
Tier 330 | -
Tier 445 | -
Tier 560 | -
Tier 690 | -
gpt-image-1, other limits
MetricLimit
Default GPT-image-1 quota limits (quick reference)9 requests per minute

gpt-image-1-mini

6 tiers - Quota Tier 0-6

gpt-image-1-mini limits
TierGlobalStandard RPM | TPM
Tier 112 | -
Tier 227 | -
Tier 354 | -
Tier 484 | -
Tier 5120 | -
Tier 6180 | -
gpt-image-1-mini, other limits
MetricLimit
Default GPT-image-1-mini quota limits (quick reference)12 requests per minute

gpt-image-1.5

6 tiers - Quota Tier 0-6

gpt-image-1.5 limits
TierDataZoneStandard RPM | TPM
Tier 13 | -
GlobalStandard RPM | TPM
9 | -
Tier 25 | -
GlobalStandard RPM | TPM
15 | -
Tier 310 | -
GlobalStandard RPM | TPM
30 | -
Tier 415 | -
GlobalStandard RPM | TPM
45 | -
Tier 520 | -
GlobalStandard RPM | TPM
60 | -
Tier 630 | -
GlobalStandard RPM | TPM
90 | -
gpt-image-1.5, other limits
MetricLimit
Default GPT-image-1.5 quota limits (quick reference)9 requests per minute

gpt-image-2

6 tiers - Quota Tier 0-6

gpt-image-2 limits
TierDataZoneStandard RPM | TPM
Tier 12 | -
GlobalStandard RPM | TPM
6 | -
Tier 24 | -
GlobalStandard RPM | TPM
12 | -
Tier 36 | -
GlobalStandard RPM | TPM
18 | -
Tier 48 | -
GlobalStandard RPM | TPM
24 | -
Tier 510 | -
GlobalStandard RPM | TPM
30 | -
Tier 612 | -
GlobalStandard RPM | TPM
36 | -
gpt-image-2, other limits
MetricLimit
Default GPT-image-2 quota limits (quick reference)9 requests per minute
  • The quick-reference default (9 requests per minute) differs from the Tier 1 GlobalStandard table value (6); both are published on the same page.

gpt-image-2.5-flare

6 tiers - Quota Tier 0-6

gpt-image-2.5-flare limits
TierGlobalStandard RPM | TPM
Tier 15 | -
Tier 25 | -
Tier 35 | -
Tier 45 | -
Tier 55 | -
Tier 65 | -

gpt-image-2.5-sunburst

6 tiers - Quota Tier 0-6

gpt-image-2.5-sunburst limits
TierGlobalStandard RPM | TPM
Tier 15 | -
Tier 25 | -
Tier 35 | -
Tier 45 | -
Tier 55 | -
Tier 65 | -

gpt-realtime

6 tiers - Quota Tier 0-6

gpt-realtime limits
TierGlobalStandard RPM | TPM
Tier 1200 | 100,000
Tier 2200 | 100,000
Tier 3200 | 100,000
Tier 4200 | 100,000
Tier 5200 | 100,000
Tier 6300 | 150,000

model-router

6 tiers - Quota Tier 0-6

model-router limits
TierDataZoneStandard RPM | TPM
Tier 1300 | 300,000
GlobalStandard RPM | TPM
1,000 | 1,000,000
Tier 2670 | 670,000
GlobalStandard RPM | TPM
2,000 | 2,000,000
Tier 31,000 | 1,000,000
GlobalStandard RPM | TPM
4,000 | 4,000,000
Tier 42,000 | 2,000,000
GlobalStandard RPM | TPM
7,000 | 7,000,000
Tier 53,000 | 3,000,000
GlobalStandard RPM | TPM
10,000 | 10,000,000
Tier 64,000 | 4,000,000
GlobalStandard RPM | TPM
15,000 | 15,000,000

o1

6 tiers - Quota Tier 0-6Model page: o1

o1 limits
TierDataZoneStandard RPM | TPM
Tier 1100 | 600,000
GlobalStandard RPM | TPM
500 | 3,000,000
Tier 2225 | 1,350,000
GlobalStandard RPM | TPM
1,000 | 6,000,000
Tier 3450 | 2,700,000
GlobalStandard RPM | TPM
2,000 | 12,000,000
Tier 4700 | 4,200,000
GlobalStandard RPM | TPM
4,000 | 24,000,000
Tier 51,000 | 6,000,000
GlobalStandard RPM | TPM
5,000 | 30,000,000
Tier 62,000 | 12,000,000
GlobalStandard RPM | TPM
8,000 | 48,000,000
o1, other limits
MetricLimit
Usage tier per month (latency threshold)4 billion tokens

o3

6 tiers - Quota Tier 0-6Model page: o3

o3 limits
TierDataZoneStandard RPM | TPM
Tier 1300 | 300,000
GlobalStandard RPM | TPM
1,000 | 1,000,000
Tier 2670 | 670,000
GlobalStandard RPM | TPM
2,000 | 2,000,000
Tier 31,000 | 1,000,000
GlobalStandard RPM | TPM
4,000 | 4,000,000
Tier 42,000 | 2,000,000
GlobalStandard RPM | TPM
7,000 | 7,000,000
Tier 53,000 | 3,000,000
GlobalStandard RPM | TPM
10,000 | 10,000,000
Tier 64,000 | 4,000,000
GlobalStandard RPM | TPM
15,000 | 15,000,000
o3, other limits
MetricLimit
Usage tier per month (latency threshold)5 billion tokens

o3-deep-research

6 tiers - Quota Tier 0-6

o3-deep-research limits
TierGlobalStandard RPM | TPM
Tier 13,000 | 3,000,000
Tier 27,000 | 7,000,000
Tier 313,000 | 13,000,000
Tier 421,000 | 21,000,000
Tier 530,000 | 30,000,000
Tier 645,000 | 45,000,000

o3-mini

6 tiers - Quota Tier 0-6Model page: o3-mini

o3-mini limits
TierDataZoneStandard RPM | TPM
Tier 1200 | 2,000,000
GlobalStandard RPM | TPM
500 | 5,000,000
Tier 2350 | 3,500,000
GlobalStandard RPM | TPM
1,000 | 10,000,000
Tier 3900 | 9,000,000
GlobalStandard RPM | TPM
2,000 | 20,000,000
Tier 41,000 | 10,000,000
GlobalStandard RPM | TPM
4,000 | 40,000,000
Tier 52,000 | 20,000,000
GlobalStandard RPM | TPM
5,000 | 50,000,000
Tier 63,000 | 30,000,000
GlobalStandard RPM | TPM
8,000 | 80,000,000
o3-mini, other limits
MetricLimit
Usage tier per month (latency threshold)50 billion tokens
Global batch enqueued tokens (Enterprise and MCA-E / Default / Credit card / MSDN / Students, trials)15B / 1B / 50M / 90K / N/A
Data zone batch enqueued tokens (Enterprise and MCA-E / Default / Credit card / MSDN / Students, trials)1.5B / 100M / 50M / 90K / N/A

o3-pro

6 tiers - Quota Tier 0-6Model page: o3-pro

o3-pro limits
TierGlobalStandard RPM | TPM
Tier 1160 | 1,600,000
Tier 2350 | 3,500,000
Tier 3715 | 7,150,000
Tier 41,150 | 11,500,000
Tier 51,600 | 16,000,000
Tier 62,400 | 24,000,000

o4-mini

7 tiers - Quota Tier 0-6Model page: o4-mini

o4-mini limits
TierGlobalStandard RPM | TPM
Tier 0100 | 100,000
Tier 11,000 | 1,000,000
DataZoneStandard RPM | TPM
300 / 10s | 300,000
Tier 22,000 | 2,000,000
DataZoneStandard RPM | TPM
670 / 10s | 670,000
Tier 34,000 | 4,000,000
DataZoneStandard RPM | TPM
1000 / 10s | 1,000,000
Tier 47,000 | 7,000,000
DataZoneStandard RPM | TPM
2000 / 10s | 2,000,000
Tier 510,000 | 10,000,000
DataZoneStandard RPM | TPM
3000 / 10s | 3,000,000
Tier 615,000 | 15,000,000
DataZoneStandard RPM | TPM
4000 / 10s | 4,000,000
o4-mini, other limits
MetricLimit
Usage tier per month (latency threshold)50 billion tokens
Global batch enqueued tokens (Enterprise and MCA-E / Default / Credit card / MSDN / Students, trials)15B / 1B / 50M / 90K / N/A
  • One of four models with a published Tier 0 (Free Tier) row.

text-embedding-3-large

6 tiers - Quota Tier 0-6

text-embedding-3-large limits
TierDataZoneStandard RPM | TPM
Tier 11,000 | 1,000,000
GlobalStandard RPM | TPM
1000 / 10s | 1,000,000
Tier 22,000 | 2,000,000
GlobalStandard RPM | TPM
2000 / 10s | 2,000,000
Tier 34,000 | 4,000,000
GlobalStandard RPM | TPM
4000 / 10s | 4,000,000
Tier 47,000 | 7,000,000
GlobalStandard RPM | TPM
7000 / 10s | 7,000,000
Tier 510,000 | 10,000,000
GlobalStandard RPM | TPM
10000 / 10s | 10,000,000
Tier 615,000 | 15,000,000
GlobalStandard RPM | TPM
15000 / 10s | 15,000,000

text-embedding-3-small

7 tiers - Quota Tier 0-6

text-embedding-3-small limits
TierGlobalStandard RPM | TPM
Tier 01000 / 10s | 1,000,000
Tier 11000 / 10s | 1,000,000
DataZoneStandard RPM | TPM
1,000 | 1,000,000
Tier 22000 / 10s | 2,000,000
DataZoneStandard RPM | TPM
2,000 | 2,000,000
Tier 34000 / 10s | 4,000,000
DataZoneStandard RPM | TPM
4,000 | 4,000,000
Tier 47000 / 10s | 7,000,000
DataZoneStandard RPM | TPM
7,000 | 7,000,000
Tier 510000 / 10s | 10,000,000
DataZoneStandard RPM | TPM
10,000 | 10,000,000
Tier 615000 / 10s | 15,000,000
DataZoneStandard RPM | TPM
15,000 | 15,000,000
  • One of four models with a published Tier 0 (Free Tier) row.

GPT-Live API

GPT-Live API limits
MetricLimit
Maximum concurrent sessions - Default (non-tiered)10
Maximum concurrent sessions - Tier 125
Maximum concurrent sessions - Tier 250
Maximum concurrent sessions - Tier 3200
Maximum concurrent sessions - Tier 4300
Maximum concurrent sessions - Tier 5500
  • 'Rate limits for the GPT-Live API are the maximum number of concurrent sessions per subscription, based on your tier.' No Tier 6 row is published.

Azure OpenAI resource and API limits (quick reference)

Azure OpenAI resource and API limits (quick reference) limits
MetricLimit
Azure OpenAI resources per region, per Azure subscription30.
Default GPT-image-1 quota limits9 requests per minute
Default GPT-image-1-mini quota limits12 requests per minute
Default GPT-image-1.5 quota limits9 requests per minute
Default GPT-image-2 quota limits9 requests per minute
Default Sora quota limits60 requests per minute.
Default Sora 2 quota limits2 job requests per minute
Default speech-to-text audio API quota limits3 requests per minute.
Maximum prompt tokens per requestVaries per model.
Maximum standard deployments per resource32.
Maximum fine-tuned model deployments10.
Total number of training jobs per resource100.
Maximum simultaneously running training jobs per resourceStandard and global training: 3; Developer training: 5
Maximum training jobs queued20.
Maximum files per resource (fine-tuning)100.
Total size of all files per resource (fine-tuning)1 GB.
Maximum training job time (job fails if exceeded)720 hours.
Maximum training job size (tokens in training file) x (# of epochs)2 billion.
Maximum size of all files per upload (Azure OpenAI on your data)16 MB.
Maximum number of inputs in array with /embeddings2,048.
Maximum tokens per /embeddings request (sum across all inputs)300,000.
Maximum number of /chat/completions messages2,048.
Maximum number of /chat/completions functions128.
Maximum number of /chat/completions tools128.
Maximum number of provisioned throughput units per deployment100,000.
Maximum files per assistant or thread10,000 when using the API or the Microsoft Foundry portal.
Maximum file size for assistants and fine-tuning512 MB via the API; 200 MB via the Foundry portal.
Maximum file upload requests per resource30 requests per second.
Maximum size for all uploaded files for assistants200 GB.
Assistants token limit2,000,000 token limit.
GPT-4o and GPT-4.1 maximum images per request50.
GPT-4 vision-preview and GPT-4 turbo-2024-04-09 default maximum tokens16. GPT-4o maximum tokens defaults to 4,096.
Maximum number of custom headers in API requests10.
Message character limit1,048,576.
Message size for audio files20 MB.
  • 'The Sora 2 RPM quota only counts video job requests. Other types of requests aren't rate-limited.' More than 10 custom headers results in HTTP 431 errors. 'Quota limits are subject to change.' 429 (Too Many Requests) can occur even when token usage metrics appear below quota.

Azure OpenAI Batch API (input files)

Azure OpenAI Batch API (input files) limits
MetricLimit
Maximum batch input files (no expiration)500
Maximum batch input files (expiration set)10,000
Maximum input file size200 MB
Maximum input file size - Bring your own storage (BYOS)1 GB
Maximum requests per file100,000
  • Batch input file limits don't apply to output files (result.jsonl, error.jsonl). Global batch quota is counted in enqueued tokens until the job reaches a terminal state. Usage tiers don't apply to global batch and provisioned throughput deployments. Per-model enqueued-token quotas are attached to each model record.

gpt-4-turbo

Global batch by subscription offer typeModel page: GPT-4 Turbo

gpt-4-turbo limits
MetricGlobal batch by subscription offer type
Global batch enqueued tokens - Enterprise and MCA-E300M
Global batch enqueued tokens - Default80M
Global batch enqueued tokens - Monthly credit card-based subscriptions40M
Global batch enqueued tokens - MSDN subscriptions90K
Global batch enqueued tokens - Azure for Students, free trialsN/A
  • B = billion | M = million | K = thousand. Not in the quota-tier RPM/TPM tables.

gpt-4 (usage tier covers gpt-4 + gpt-4-32k, all versions)

Global batch by subscription offer type; usage tierModel page: GPT-4

gpt-4 (usage tier covers gpt-4 + gpt-4-32k, all versions) limits
MetricGlobal batch by subscription offer type; usage tier
Global batch enqueued tokens - Enterprise and MCA-E150M
Global batch enqueued tokens - Default30M
Global batch enqueued tokens - Monthly credit card-based subscriptions5M
Global batch enqueued tokens - MSDN subscriptions100K
Global batch enqueued tokens - Azure for Students, free trialsN/A
Usage tier per month (gpt-4 + gpt-4-32k, all versions)6 billion tokens

Usage tiers (Global Standard, Data Zone Standard, Standard)

Usage tiers (Global Standard, Data Zone Standard, Standard) limits
MetricLimit
Usage tier definitionUsage tiers per month, in tokens per model; usage is the total tokens consumed across all deployments in all subscriptions in all regions for a given tenant
  • Exceeding the usage tier is not a hard stop: 'Latency can vary and, in some cases, may be more than two times higher than when operating within your usage tier.' Usage tiers don't apply to global batch and provisioned throughput deployments. Per-model monthly values are attached to each model record.

gpt-oss-120b

gpt-oss-120b limits
MetricLimit
TPM5 M
RPM5 K

Grok 4.7 version 1 (Foundry Models)

7 tiers - Subscription tier: Free Tier, Tier 1-6

Grok 4.7 version 1 (Foundry Models) limits
TierGlobal Standard TPM | RPM
Free Tier0 | 0
Data Zone Standard (US) TPM | RPM
0 | 0
Tier 180,000 | 80
Data Zone Standard (US) TPM | RPM
80,000 | 80
Tier 2315,000 | 315
Data Zone Standard (US) TPM | RPM
315,000 | 315
Tier 31,000,000 | 1,000
Data Zone Standard (US) TPM | RPM
1,000,000 | 1,000
Tier 42,000,000 | 2,000
Data Zone Standard (US) TPM | RPM
2,000,000 | 2,000
Tier 510,000,000 | 10,000
Data Zone Standard (US) TPM | RPM
10,000,000 | 10,000
Tier 615,000,000 | 15,000
Data Zone Standard (US) TPM | RPM
15,000,000 | 15,000
  • 'Each unit of deployment capacity provides 1,000 TPM and 1 RPM.' Uses subscription-level quota management. When a model is upgraded, its new Global or Data Zone quota limit is the greater of the applicable tier limit or the total quota assigned to existing deployments in scope.

Microsoft-Decision-1 (Foundry Models)

6 tiers - Subscription tier: Tier 1-6

Microsoft-Decision-1 (Foundry Models) limits
TierRPM (Global Standard and Data Zone Standard)
Tier 160
Tier 2150
Tier 3225
Tier 4360
Tier 5500
Tier 6900
  • Global Standard and Data Zone Standard rows publish identical values.

Foundry Models sold by Azure not onboarded to subscription-level quota (per region, per subscription)

Foundry Models sold by Azure not onboarded to subscription-level quota (per region, per subscription) limits
MetricLimit
Llama 3.3 70B Instruct, Llama-4-Maverick-17B-128E-Instruct-FP8 - TPM | RPM | Concurrent requests400,000 | 1,000 | 300
Flux.2-Pro - RPMLow (Default): 15; Medium: 30; High (Enterprise): 100
FLUX-1.1-pro, Flux.1-Kontext Pro - RPM2 capacity units (6 requests per minute)
Rest of models - TPM | RPM | Concurrent requests400,000 | 1,000 | 300
Foundry resources per region per Azure subscription100
Max projects per resource250
Max deployments per resource32
Max number of custom headers in API requests10
Client-side timeout guidance - reasoning modelsup to 29 minutes
Client-side timeout guidance - non-reasoning modelsstreaming up to 60 seconds; non-streaming up to 29 minutes
  • Azure OpenAI models: 'Varies per model and SKU.' HTTP 429 Too Many Requests: implement retry logic with exponential backoff and use the Retry-After header value. HTTP 431 Request Header Fields Too Large: more than 10 custom headers. Page last updated 10/06/2026.

How to raise your limits

  1. Azure OpenAI in Foundry Models (Global Standard, Data Zone Standard, Standard), codex-mini and 61 more

    Automatic: 'If a customer's usage increases such that their current quota tier limits their ability to use Foundry Models, the system automatically upgrades the customer to the next higher tier.' Enterprise relationships (EA, MCA-E) are assigned higher tiers; payment history is considered for automatic upgrades. Initial tier is based on current usage of that model and relationship with Microsoft. Opt out (preview): PATCH https://management.azure.com/subscriptions/{id}/providers/Microsoft.CognitiveServices/quotaTiers/default?api-version=2025-10-01-preview with {"properties": {"tierUpgradePolicy": "NoAutoUpgrade"}}. Check tier: GET https://management.azure.com/subscriptions/{id}/providers/Microsoft.CognitiveServices/quotaTiers?api-version=2025-10-01-preview (returns currentTierName, tierUpgradePolicy such as OnceUpgradeIsAvailable). Extra quota: quota request form https://aka.ms/oai/stuquotarequest; if approved the current tier stays the same with more quota assigned. 'Quota increase requests are processed in the order they're received, and priority goes to customers who actively use their existing quota allocation.'

  2. GPT-Live API

    Tier-based; tiers advance automatically with usage.

  3. Azure OpenAI resource and API limits (quick reference), gpt-4-turbo and 2 more

    Submit the quota increase request form https://aka.ms/oai/stuquotarequest. 'Quota increase requests are processed in the order they're received, and priority goes to customers who actively use their existing quota allocation. Requests that don't meet this condition might be denied.'

  4. Azure OpenAI Batch API (input files)

    Set expiration for input files to raise the per-resource file limit; use Batch with Azure Blob Storage to avoid Files API input file limits. Batch enqueued-token quota increases via https://aka.ms/oai/stuquotarequest.

  5. Usage tiers (Global Standard, Data Zone Standard, Standard)

    If you exceed your usage tier: request a quota increase in the Azure portal, consider upgrading to Provisioned Throughput Units (PTU), and monitor usage.

  6. gpt-4-turbo, gpt-4 (usage tier covers gpt-4 + gpt-4-32k, all versions) and 2 more

    Quota request form https://aka.ms/oai/stuquotarequest ('requests to increase quota are evaluated individually'). Except for Anthropic models, Models from partners and community don't support quota increases.

What the docs leave out

  • Provisioned throughput (PTU) per-model capacity, minimums and throughput-per-PTU are not on the quotas page (only 'Maximum number of provisioned throughput units per deployment: 100,000.').
  • Current assigned tier, per-region Standard quota availability and remaining capacity are only visible in the logged-in Foundry portal (https://ai.azure.com/resource/quota) or via authenticated Usages, Model Capacities and quotaTiers APIs.
  • No numeric thresholds are published for automatic tier upgrades (criteria are described qualitatively: consumption trends, Enterprise relationship, payment history).
  • No Tier 6 row is published for GPT-Live concurrent sessions.
  • Tier 0 (Free Tier) rows are published only for gpt-4.1-mini, gpt-5-mini, o4-mini and text-embedding-3-small.
  • Anthropic model quotas in Foundry are not tabulated on these pages (only that they support quota increase requests).
  • Sora / Sora 2 and speech-to-text have only the quick-reference default RPM; no per-tier values published.

FAQ

What are the Azure OpenAI quotas and limits?

Microsoft's official docs, read on October 17, 2026, set quotas and limits for Azure OpenAI in Foundry. For codex-mini, Tier 1 is listed at 1,000 RPM and 1,000,000 TPM (GlobalStandard), and Tier 6 at 15,000 RPM and 15,000,000 TPM (GlobalStandard).

How do I raise my Azure OpenAI in Foundry quotas and limits?

Automatic: 'If a customer's usage increases such that their current quota tier limits their ability to use Foundry Models, the system automatically upgrades the customer to the next higher tier.' Enterprise relationships (EA, MCA-E) are assigned higher tiers; payment history is considered for automatic upgrades. Initial tier is based on current usage of that model and relationship with Microsoft. Opt out (preview): PATCH https://management.azure.com/subscriptions/{id}/providers/Microsoft.CognitiveServices/quotaTiers/default?api-version=2025-10-01-preview with {"properties": {"tierUpgradePolicy": "NoAutoUpgrade"}}. Check tier: GET https://management.azure.com/subscriptions/{id}/providers/Microsoft.CognitiveServices/quotaTiers?api-version=2025-10-01-preview (returns currentTierName, tierUpgradePolicy such as OnceUpgradeIsAvailable). Extra quota: quota request form https://aka.ms/oai/stuquotarequest; if approved the current tier stays the same with more quota assigned. 'Quota increase requests are processed in the order they're received, and priority goes to customers who actively use their existing quota allocation.' Tier-based; tiers advance automatically with usage. Submit the quota increase request form https://aka.ms/oai/stuquotarequest. 'Quota increase requests are processed in the order they're received, and priority goes to customers who actively use their existing quota allocation. Requests that don't meet this condition might be denied.'

Where do these Azure OpenAI numbers come from?

Every value is copied verbatim from Microsoft's official documentation at learn.microsoft.com, read on October 17, 2026. This page lists all 68 entries in 68 tables, each linked to the page it came from.

Related reading

Sources

Last verified . Every value is copied verbatim from the official pages below, read on that date.

Newsletter

Know when limits change

One email per week. New model releases, pricing and limit changes, and what they mean for what you build.