Microsoft Azure: Models Intelligence, Performance & Price

Name: Intelligence
Creator: Artificial Analysis
License: https://artificialanalysis.ai/docs/legal/Terms-of-Use.pdf

Microsoft Azure

Analysis of Microsoft Azure's models across key metrics including quality, price, output speed, latency, context window & more. This analysis is intended to support you in choosing the best model provided by Microsoft Azure for your use-case.

Most Intelligent

Updated

Claude Fable 5 (with fallback)

Claude Opus 4.7 (max)

Claude Sonnet 5 (max)

GPT-5.4 (xhigh)

Claude Sonnet 4.6 (max)

Intelligence index

Total 79 models

Fastest

gpt-oss-120b (low)

343 t/s

gpt-oss-120b (high)

328 t/s

GPT-4.1 nano

259 t/s

GPT-5.4 mini (xhigh)

241 t/s

o3-mini

219 t/s

Output speed

Total 79 models

Lowest Price

GPT-5 nano (high)

$0.06

GPT-5 nano (medium)

$0.06

GPT-4.1 nano

$0.08

GPT-4o mini

$0.14

Phi-4

$0.16

Blended price (per 1M tokens)

Total 79 models

Indicates a reasoning model

Azure offers 79 models, each with different intelligence, performance, and pricing characteristics. Below is a comparison of the key metrics across models.

For intelligence, the top models on Azure are Claude Fable 5 (with fallback) (60), Claude Opus 4.7 (max) (54), Claude Sonnet 5 (max) (53).
For output speed, the fastest models are gpt-oss-120b (low) (343 t/s), gpt-oss-120b (high) (328 t/s), GPT-4.1 nano (259 t/s). Speed varies significantly across models, with a 57% difference between the fastest and slowest.
For latency, gpt-oss-120b (high) (0.67s), gpt-oss-120b (low) (0.68s), Phi-4 Mini (0.83s) offer the lowest time to first token.
For pricing, GPT-5 nano (high) ($0.06), GPT-5 nano (medium) ($0.06), GPT-4.1 nano ($0.08) offer the lowest blended prices per 1M tokens. Prices vary up to 2.9x across models.
For context window size, GPT-5.4 (xhigh) (1M), Claude Fable 5 (with fallback) (1M), Claude Opus 4.7 (max) (1M) support the largest context windows on Azure.

Highlights

Updated

Intelligence

Artificial Analysis Intelligence Index · Higher is better

Speed

Output tokens per second · Higher is better

Price

USD per 1M tokens (blended) · Lower is better

Intelligence Evaluations

Artificial Analysis Intelligence Index

Artificial Analysis Intelligence Index v4.1 incorporates 9 evaluations: GDPval-AA v2, 𝜏³-Banking, Terminal-Bench v2.1, SciCode, Humanity's Last Exam, GPQA Diamond, CritPt, AA-Omniscience, AA-LCR

Estimate (independent evaluation forthcoming)

Reasoning models are indicated by a lightbulb icon

Artificial Analysis Intelligence Index v4.1 includes: GDPval-AA v2, 𝜏³-Banking, Terminal-Bench v2.1, SciCode, Humanity's Last Exam, GPQA Diamond, CritPt, AA-Omniscience, AA-LCR. See Intelligence Index methodology for further details, including a breakdown of each evaluation and how we run them.

Intelligence Evaluations

Intelligence evaluations measured independently by Artificial Analysis · Higher is better

GDPval-AA v2

Agentic real-world work tasks, (Elo-500)/2000

𝜏³-Banking

Agentic tool use

Terminal-Bench v2.1

Agentic coding & terminal use

SciCode

Coding

Humanity's Last Exam

Reasoning & knowledge

GPQA Diamond

Scientific reasoning

CritPt

Physics reasoning

AA-Omniscience Accuracy

Knowledge

AA-Omniscience Non-Hallucination Rate

1 - hallucination rate

AA-LCR

Long context reasoning

AA-BriefcaseNew

Agentic knowledge work, Elo

AutomationBench-AANew

Agentic SaaS workflows

Harvey LAB-AANew

Legal agentic work, task all-pass rate

EnterpriseOps-Gym-AANew

Agentic business operations

IFBench

Instruction following

APEX-Agents-AA

Long-horizon agentic tasks

ITBench-AA

Kubernetes incident root-cause analysis

MMMU-Pro

Visual reasoning

Reasoning models are indicated by a lightbulb icon.

While model intelligence generally translates across use cases, specific evaluations may be more relevant for certain use cases.

Intelligence vs. Price

Blended at 7:2:1 (cache-input-output) · USD per 1M tokens (blended)

Most attractive quadrant

Reasoning models are indicated by a lightbulb icon.

While higher intelligence models are typically more expensive, they do not all follow the same price-quality curve.

Price per token, shown in USD per million tokens. Price is a blend of cache hit, input, and output token prices using the selected ratio (default 7:2:1 cache-input-output).

The blended cache price shown here uses cache hit price only. Other caching costs differ by provider:

Anthropic: charges a separate cache write fee, with different rates for 5-minute and 1-hour TTLs (1-hour TTL is more expensive).
Google (Vertex/Gemini): charges a per-hour cache storage fee in addition to cache hit pricing. Some providers also use tiered pricing for prompts above 200K tokens.
OpenAI, DeepSeek, others: typically charge only cache hit pricing with no write or storage fee.

See Prompt Caching for the full breakdown.

Figures represent performance of the model's first-party API (e.g. OpenAI for o1) or the median across providers where a first-party API is not available (e.g. Meta's Llama models).

Context Window

Context window: tokens limit · Higher is better

Reasoning models are indicated by a lightbulb icon

Larger context windows are relevant to RAG (Retrieval Augmented Generation) LLM workflows which typically involve reasoning and information retrieval of large amounts of data.

Maximum number of combined input & output tokens. Output tokens commonly have a significantly lower limit (varied by model).

Pricing

Intelligence vs. Price

Blended at 7:2:1 (cache-input-output) · USD per 1M tokens (blended)

Most attractive quadrant

Reasoning models are indicated by a lightbulb icon.

While higher intelligence models are typically more expensive, they do not all follow the same price-quality curve.

Price per token, shown in USD per million tokens. Price is a blend of cache hit, input, and output token prices using the selected ratio (default 7:2:1 cache-input-output).

The blended cache price shown here uses cache hit price only. Other caching costs differ by provider:

Anthropic: charges a separate cache write fee, with different rates for 5-minute and 1-hour TTLs (1-hour TTL is more expensive).
Google (Vertex/Gemini): charges a per-hour cache storage fee in addition to cache hit pricing. Some providers also use tiered pricing for prompts above 200K tokens.
OpenAI, DeepSeek, others: typically charge only cache hit pricing with no write or storage fee.

See Prompt Caching for the full breakdown.

Figures represent performance of the model's first-party API (e.g. OpenAI for o1) or the median across providers where a first-party API is not available (e.g. Meta's Llama models).

Performance Summary

Output Speed vs. Price

Output speed: output tokens per second · USD per 1M tokens (blended)

Most attractive quadrant

Reasoning models are indicated by a lightbulb icon.

Tokens per second received while the model is generating tokens (ie. after first chunk has been received from the API for models which support streaming).

Price per token, shown in USD per million tokens. Price is a blend of cache hit, input, and output token prices using the selected ratio (default 7:2:1 cache-input-output).

Speed

Measured by Output Speed (tokens per second)

Output Speed

Output tokens per second · Higher is better

Reasoning models are indicated by a lightbulb icon

Tokens per second received while the model is generating tokens (ie. after first chunk has been received from the API for models which support streaming).

Figures represent performance of the model's first-party API (e.g. OpenAI for o1) or the median across providers where a first-party API is not available (e.g. Meta's Llama models).

Latency

Measured by Time (seconds) to First Token

Latency: Time To First Answer Token

Seconds to first answer token received · Accounts for reasoning model 'thinking' time

Reasoning models are indicated by a lightbulb icon

Time to first answer token received, in seconds, after API request sent. For reasoning models, this includes the 'thinking' time of the model before providing an answer. For models which do not support streaming, this represents time to receive the completion.

End-to-End Response Time

Seconds to output 500 tokens, calculated based on time to first token, 'thinking' time for reasoning models, and output speed

End-to-End Response Time vs. Price

End-to-end response time: end-to-end seconds to output 500 tokens · USD per 1M tokens (blended)

Most attractive quadrant

Reasoning models are indicated by a lightbulb icon.

Price per token, shown in USD per million tokens. Price is a blend of cache hit, input, and output token prices using the selected ratio (default 7:2:1 cache-input-output).

Seconds to receive a 500 token response. Key components:

Input time: Time to receive the first response token
Thinking time (only for reasoning models): Time reasoning models spend outputting tokens to reason prior to providing an answer. Amount of tokens based on the average reasoning tokens across a diverse set of 60 prompts (methodology details).
Answer time: Time to generate 500 output tokens, based on output speed

Figures represent median (P50) measurement over the past 72 hours to reflect sustained changes in performance.


									Further Analysis
Claude Fable 5 (with fallback)	1M	Proprietary	60	$14.00	65	98.71	106.42	--	Model Provider
Claude Opus 4.7 (max)	1M	Proprietary	54	$3.85	47	32.69	43.25	--	Model Provider
Claude Sonnet 5 (max)	1M	Proprietary	53	$2.31	80	143.78	150.04	--	Model Provider
GPT-5.4 (xhigh)	1.05M	Proprietary	51	$2.17	97	146.48	151.64	--	Model Provider
Claude Sonnet 4.6 (max)	200k	Proprietary	47	$2.31	45	114.55	125.70	--	Model Provider
DeepSeek V4 Pro (max)	1M	Open	44	$1.91	68	1.57	73.13	64.22	Model Provider
Kimi K2.6	262k	Open	44	$1.25	186	1.60	28.27	23.98	Model Provider
Claude Opus 4.6 (max)	1M	Proprietary	44*	$3.85	40	15.62	28.15	--	Model Provider
GPT-5.2 (xhigh)	400k	Proprietary	42*	$1.87	78	187.46	193.88	--	Model Provider
DeepSeek V4 Pro (high)	1M	Open	41*	$1.91	68	1.64	38.55	29.50	Model Provider
Claude Opus 4.5	200k	Proprietary	41*	$7.00	47	10.07	20.78	--	Model Provider
GPT-5.2 Codex (xhigh)	400k	Proprietary	40*	$1.88	127	124.82	128.75	--	Model Provider
GPT-5.4 mini (xhigh)	400k	Proprietary	40	$0.65	241	142.49	144.57	--	Model Provider
GPT-5.2 (medium)	400k	Proprietary	38*	$2.98	71	8.11	15.20	--	Model Provider
Claude Opus 4.6 (high)	1M	Proprietary	38*	$3.85	37	2.33	15.92	--	Model Provider
Grok 4.3 (high)	200k	Proprietary	38	$1.38	105	27.64	32.41	--	Model Provider
Grok 4.20 0309 v2	262k	Proprietary	37*	$2.40	131	16.15	19.97	--	Model Provider
GPT-5.1 (high)	272k	Proprietary	37	$1.34	114	42.15	46.54	--	Model Provider
Claude 4.5 Sonnet	200k	Proprietary	36	$4.20	41	11.12	23.32	--	Model Provider
GPT-5 Codex (high)	400k	Proprietary	36*	$1.34	136	18.81	22.49	--	Model Provider
Grok 4.3 (medium)	200k	Proprietary	36*	$1.38	112	17.01	21.50	--	Model Provider
Claude Sonnet 4.6 (Non-reasoning)	200k	Proprietary	36*	$2.31	41	1.76	13.83	--	Model Provider
Grok 4.3 (low)	200k	Proprietary	35*	$1.38	109	6.77	11.36	--	Model Provider
Kimi K2.5	262k	Open	35	$0.84	194	1.32	19.17	15.27	Model Provider
GPT-5.1 Codex (high)	400k	Proprietary	35*	$2.13	120	8.04	12.20	--	Model Provider
GPT-5 (high)	400k	Proprietary	35	$1.34	92	65.87	71.33	--	Model Provider
Claude Opus 4.5	200k	Proprietary	35*	$7.00	46	2.93	13.82	--	Model Provider
Kimi K2.6	262k	Open	35*	$1.25	185	1.63	4.33	--	Model Provider
GPT-5 (medium)	400k	Proprietary	34*	$1.34	79	25.83	32.15	--	Model Provider
Claude 4.1 Opus	200k	Proprietary	34*	$21.00	30	11.76	28.58	--	Model Provider
Grok 4	256k	Proprietary	33*	$7.70	45	16.69	27.75	--	Model Provider
Kimi K2 Thinking	256k	Open	33*	$0.79	184	1.30	14.86	10.85	Model Provider
o3-pro	200k	Proprietary	33*	$26.00	25	168.02	188.42	--	Model Provider
DeepSeek V4 Pro	1M	Open	31*	$1.91	63	1.67	9.62	--	Model Provider
GPT-5 (low)	400k	Proprietary	31*	$1.34	79	8.10	14.47	--	Model Provider
GPT-5 mini (medium)	400k	Proprietary	31*	$0.27	100	11.39	16.41	--	Model Provider
GPT-5.1 Codex mini (high)	400k	Proprietary	31*	$0.42	154	29.52	32.77	--	Model Provider
o3	200k	Proprietary	30*	$1.55	106	31.09	35.78	--	Model Provider
GPT-5.4 mini (medium)	400k	Proprietary	30*	$1.13	179	11.50	14.29	--	Model Provider
Claude 4.5 Haiku	200k	Proprietary	30	$1.40	91	20.23	25.73	--	Model Provider
Kimi K2.5	262k	Open	29*	$0.84	162	1.30	4.39	--	Model Provider
Claude 4.5 Sonnet	200k	Proprietary	29*	$4.20	41	2.07	14.38	--	Model Provider
Claude 4.1 Opus	200k	Proprietary	28*	$21.00	30	2.86	19.52	--	Model Provider
GPT-5.2	400k	Proprietary	26*	$2.98	69	1.71	8.96	--	Model Provider
o4-mini (high)	200k	Proprietary	26*	$0.86	138	22.34	25.97	--	Model Provider
GPT-5 mini (high)	400k	Proprietary	25	$0.27	103	88.02	92.89	--	Model Provider
DeepSeek V3.2	128k	Open	25*	$0.69	215	1.38	3.71	--	Model Provider
gpt-oss-120b (high)	131k	Open	24	$0.20	328	0.67	8.28	6.09	Model Provider
Claude 4.5 Haiku	200k	Proprietary	24*	$1.40	88	1.32	7.01	--	Model Provider
o1	200k	Proprietary	23*	$14.25	112	44.25	48.70	--	Model Provider
GPT-5.1	400k	Proprietary	20*	$2.13	104	1.44	6.24	--	Model Provider
DeepSeek R1 0528	128k	Open	20*	$1.76	86	1.78	30.97	23.35	Model Provider
GPT-5 nano (high)	400k	Proprietary	20*	$0.06	175	94.57	97.42	--	Model Provider
GPT-4.1	1M	Proprietary	19*	$1.55	86	1.66	7.50	--	Model Provider
GPT-5 nano (medium)	400k	Proprietary	19*	$0.06	177	35.90	38.72	--	Model Provider
o3-mini	200k	Proprietary	19*	$1.05	216	8.49	10.81	--	Model Provider
DeepSeek R1 (Jan)	128k	Open	19	$1.76	87	1.83	34.63	27.04	Model Provider
GPT-5 (minimal)	400k	Proprietary	17*	$1.34	82	1.63	7.72	--	Model Provider
o1-preview	128k	Proprietary	17*	$15.68	142	24.44	27.94	--	Model Provider
GPT-5.4 mini	400k	Proprietary	17*	$1.13	182	0.86	3.62	--	Model Provider
Mistral Large 3	256k	Open	16	$0.60	42	1.75	13.67	--	Model Provider
o3-mini (high)	200k	Proprietary	16	$1.05	200	23.77	26.28	--	Model Provider
DeepSeek V3 0324	128k	Open	15	$1.48	90	1.89	7.41	--	Model Provider
gpt-oss-120b (low)	131k	Open	15	$0.20	348	0.69	7.88	5.75	Model Provider
GPT-4.1 mini	1M	Proprietary	15	$0.31	92	1.35	6.76	--	Model Provider
Llama 4 Maverick (FP8)	128k	Open	14	$0.46	115	1.13	5.47	--	Model Provider
Mistral Medium 3	128k	Proprietary	12*	$0.56	33	2.26	17.58	--	Model Provider
GPT-4o (Nov)	128k	Proprietary	11*	$2.38	125	2.21	6.21	--	Model Provider
Llama 4 Scout	128k	Open	10	$0.26	102	0.96	5.88	--	Model Provider
GPT-4o (Aug)	128k	Proprietary	10*	$2.38	144	1.24	4.71	--	Model Provider
GPT-4.1 nano	1M	Proprietary	10	$0.08	259	1.09	3.02	--	Model Provider
Llama 3.3 70B	128k	Open	9	$0.71	89	2.46	8.07	--	Model Provider
GPT-4o (May)	128k	Proprietary	9*	$4.25	138	1.46	5.09	--	Model Provider
GPT-4 Turbo	128k	Proprietary	8*	$12.00	89	1.68	7.33	--	Model Provider
Command A	256k	Open	8*	$3.25	41	3.22	15.41	--	Model Provider
GPT-4o mini	128k	Proprietary	7*	$0.14	81	1.83	7.99	--	Model Provider
Phi-4 Mini	128k	Open	6	$0.00	45	0.83	11.90	--	Model Provider
Phi-4	16k	Open	5*	$0.16	21	2.23	26.49	--	Model Provider
Phi-4 Multimodal	128k	Open	5*	$0.00	16	0.84	31.22	--	Model Provider

Key definitions

Maximum number of combined input & output tokens. Output tokens commonly have a significantly lower limit (varied by model).

Tokens per second received while the model is generating tokens (ie. after first chunk has been received from the API for models which support streaming).

Time to first token received, in seconds, after API request sent. For reasoning models which share reasoning tokens, this will be the first reasoning token. For models which do not support streaming, this represents time to receive the completion.

Price per token, shown in USD per million tokens. Price is a blend of cache hit, input, and output token prices using the selected ratio (default 7:2:1 cache-input-output).

Price per token generated by the model (received from the API), represented as USD per million Tokens.

Price per token included in the request/message sent to the API, represented as USD per million Tokens.

Metrics are 'live' and are based on the past 72 hours of measurements, measurements are taken 8 times a day for single requests and 2 times per day for parallel requests.

Frequently Asked Questions

Common questions about Microsoft Azure

Microsoft Azure offers 79 models that we track: Claude Fable 5 (with fallback), Claude Opus 4.7 (max), Claude Sonnet 5 (max), GPT-5.4 (xhigh), Claude Sonnet 4.6 (max), DeepSeek V4 Pro (max), Kimi K2.6, Claude Opus 4.6 (max), GPT-5.2 (xhigh), DeepSeek V4 Pro (high), Claude Opus 4.5, GPT-5.2 Codex (xhigh), GPT-5.4 mini (xhigh), GPT-5.2 (medium), Claude Opus 4.6 (high), Grok 4.3 (high), Grok 4.20 0309 v2, GPT-5.1 (high), Claude 4.5 Sonnet, GPT-5 Codex (high), Grok 4.3 (medium), Claude Sonnet 4.6 (Non-reasoning), Grok 4.3 (low), Kimi K2.5, GPT-5.1 Codex (high), GPT-5 (high), Claude Opus 4.5, Kimi K2.6, GPT-5 (medium), Claude 4.1 Opus, Grok 4, Kimi K2 Thinking, o3-pro, DeepSeek V4 Pro, GPT-5 (low), GPT-5 mini (medium), GPT-5.1 Codex mini (high), o3, GPT-5.4 mini (medium), Claude 4.5 Haiku, Kimi K2.5, Claude 4.5 Sonnet, Claude 4.1 Opus, GPT-5.2, o4-mini (high), GPT-5 mini (high), DeepSeek V3.2, gpt-oss-120b (high), Claude 4.5 Haiku, o1, GPT-5.1, DeepSeek R1 0528, GPT-5 nano (high), GPT-4.1, GPT-5 nano (medium), o3-mini, DeepSeek R1 (Jan), GPT-5 (minimal), o1-preview, GPT-5.4 mini, Mistral Large 3, o3-mini (high), DeepSeek V3 0324, gpt-oss-120b (low), GPT-4.1 mini, Llama 4 Maverick (FP8), Mistral Medium 3, GPT-4o (Nov), Llama 4 Scout, GPT-4o (Aug), GPT-4.1 nano, Llama 3.3 70B, GPT-4o (May), GPT-4 Turbo, Command A, GPT-4o mini, Phi-4 Mini, Phi-4, and Phi-4 Multimodal.

The most intelligent model available on Microsoft Azure is Claude Fable 5 (with fallback) with an Intelligence Index score of 60.

The fastest model on Microsoft Azure by output speed is gpt-oss-120b (low) at 342.6 tokens per second.

The model with the lowest time to first token on Microsoft Azure is gpt-oss-120b (high) at 0.67s. Lower latency means faster initial response time.

The most affordable model on Microsoft Azure by blended price is GPT-5 nano (high) at $0.06 per 1M tokens (7:2:1 cache hit/input/output ratio).

Prices on Microsoft Azure vary up to 456x across models, from $0.06 per 1M tokens for GPT-5 nano (high) to $26.00 per 1M tokens for o3-pro.

Yes, Microsoft Azure offers an OpenAI-compatible API, making it easy to switch from OpenAI or use existing OpenAI SDK integrations.

64 of 79 models on Microsoft Azure support JSON mode for structured output.

74 of 79 models on Microsoft Azure support function calling (tool use).

Yes, Microsoft Azure offers 47 reasoning models: Claude Fable 5 (with fallback), Claude Opus 4.7 (max), Claude Sonnet 5 (max), GPT-5.4 (xhigh), Claude Sonnet 4.6 (max), DeepSeek V4 Pro (max), Kimi K2.6, Claude Opus 4.6 (max), GPT-5.2 (xhigh), DeepSeek V4 Pro (high), Claude Opus 4.5, GPT-5.2 Codex (xhigh), GPT-5.4 mini (xhigh), GPT-5.2 (medium), Grok 4.3 (high), Grok 4.20 0309 v2, GPT-5.1 (high), Claude 4.5 Sonnet, GPT-5 Codex (high), Grok 4.3 (medium), Grok 4.3 (low), Kimi K2.5, GPT-5.1 Codex (high), GPT-5 (high), GPT-5 (medium), Claude 4.1 Opus, Grok 4, Kimi K2 Thinking, o3-pro, GPT-5 (low), GPT-5 mini (medium), GPT-5.1 Codex mini (high), o3, GPT-5.4 mini (medium), Claude 4.5 Haiku, o4-mini (high), GPT-5 mini (high), gpt-oss-120b (high), o1, DeepSeek R1 0528, GPT-5 nano (high), GPT-5 nano (medium), o3-mini, DeepSeek R1 (Jan), o1-preview, o3-mini (high), and gpt-oss-120b (low). Reasoning models use extended thinking to work through complex problems before providing an answer.

Yes, 22 of 79 models on Microsoft Azure are open weight models: DeepSeek V4 Pro (max), Kimi K2.6, DeepSeek V4 Pro (high), Kimi K2.5, Kimi K2.6, Kimi K2 Thinking, DeepSeek V4 Pro, Kimi K2.5, DeepSeek V3.2, gpt-oss-120b (high), DeepSeek R1 0528, DeepSeek R1 (Jan), Mistral Large 3, DeepSeek V3 0324, gpt-oss-120b (low), Llama 4 Maverick (FP8), Llama 4 Scout, Llama 3.3 70B, Command A, Phi-4 Mini, Phi-4, and Phi-4 Multimodal.

Yes, provider performance can vary over time due to infrastructure changes, load balancing, and updates. We continuously benchmark all providers and display historical performance trends in the "Over Time" charts.

When choosing a model on Microsoft Azure, consider: intelligence (for quality-sensitive tasks), output speed (for throughput-intensive tasks), latency (for interactive applications requiring quick first responses), pricing (for cost-sensitive workloads), and features like context window size, JSON mode, or function calling support.

Microsoft Azure: Models Intelligence, Performance & Price

Most Intelligent

Fastest

Lowest Price

Intelligence