Phi-4 logo
Microsoft
•

Open weights model

•

Released December 2024

Phi-4 Intelligence, Performance & Price Analysis

API Provider Benchmarks

Model summary

Intelligence

10
Artificial Analysis Intelligence Index
2 out of 4 units for Intelligence.

Speed

35.7
Output tokens per second
1 out of 4 units for Speed.

Input PriceUpdated

$0.125

USD per 1M tokens

Cache: $0.125 (-0%)

3 out of 4 units for Input Price.

Output Price

$0.50
USD per 1M tokens
4 out of 4 units for Output Price.

Verbosity

N/A
Output tokens from Intelligence Index
Unknown out of 4 units for Verbosity.

Phi-4 is below average in intelligence and somewhat expensive when comparing to other open weight non-reasoning models of similar size. The model has a 16k tokens context window with knowledge up to June 2024.

Phi-4 scores 10 on the Artificial Analysis Intelligence Index, placing it below average among comparable models (averaging 12).

Pricing for Phi-4 is $0.13 per 1M input tokens (somewhat expensive, average: $0.05) and $0.50 per 1M output tokens (expensive, average: $0.19).

At 36 tokens per second, Phi-4 is notably slow (95).

ReasoningNo

This page shows the non-reasoning version of this model.

A reasoning variant may also exist.

Input modality

Supports:

This information is still being updated

Output modality

Supports:

This information is still being updated

Knowledge cutoffJun 1, 2024
Context window16k
~24 A4 pages of size 12 Arial font
Total parameters14B
LicenseMIT
Model weightsHugging Face

Metrics are compared against models of the same class:

  • Non-reasoning models → compared only with other non-reasoning models
  • Reasoning models → compared across both reasoning and non-reasoning
  • Open weights models → compared only with other open weights models of the same size class:
    • Tiny: ≤4B parameters
    • Small: 4B–40B parameters
    • Medium: 40B–150B parameters
    • Large: >150B parameters
  • Proprietary models → compared across proprietary and open weights models of the same price range, using a blended 3:1 input/output price ratio:
    • <$0.15 per 1M tokens
    • $0.15–$1 per 1M tokens
    • >$1 per 1M tokens

Highlights

Intelligence

Artificial Analysis Intelligence Index · Higher is better

Speed

Output tokens per second · Higher is better
Updated

Price

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

Intelligence

Artificial Analysis Intelligence Index

Artificial Analysis Intelligence Index v4.0 incorporates 10 evaluations: GDPval-AA, 𝜏²-Bench Telecom, Terminal-Bench Hard, SciCode, AA-LCR, AA-Omniscience, IFBench, Humanity's Last Exam, GPQA Diamond, CritPt
Reasoning models are indicated by a lightbulb icon

Artificial Analysis Intelligence Index v4.0 includes: GDPval-AA, 𝜏²-Bench Telecom, Terminal-Bench Hard, SciCode, AA-LCR, AA-Omniscience, IFBench, Humanity's Last Exam, GPQA Diamond, CritPt. See Intelligence Index methodology for further details, including a breakdown of each evaluation and how we run them.

Artificial Analysis Intelligence Index by Open Weights / Proprietary

Artificial Analysis Intelligence Index v4.0 incorporates 10 evaluations: GDPval-AA, 𝜏²-Bench Telecom, Terminal-Bench Hard, SciCode, AA-LCR, AA-Omniscience, IFBench, Humanity's Last Exam, GPQA Diamond, CritPt
Reasoning models are indicated by a lightbulb icon

Artificial Analysis Intelligence Index v4.0 includes: GDPval-AA, 𝜏²-Bench Telecom, Terminal-Bench Hard, SciCode, AA-LCR, AA-Omniscience, IFBench, Humanity's Last Exam, GPQA Diamond, CritPt. See Intelligence Index methodology for further details, including a breakdown of each evaluation and how we run them.

Indicates whether the model weights are available. Models are labelled as 'Commercial Use Restricted' if the weights are available but commercial use is limited (typically requires obtaining a paid license).

Intelligence Evaluations

Intelligence evaluations measured independently by Artificial Analysis · Higher is better
GDPval-AA

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

Terminal-Bench Hard

Agentic coding & terminal use

𝜏²-Bench Telecom

Agentic tool use

AA-LCR

Long context reasoning

AA-Omniscience Accuracy

Knowledge

AA-Omniscience Non-Hallucination Rate

1 - hallucination rate

Humanity's Last Exam

Reasoning & knowledge

GPQA Diamond

Scientific reasoning

SciCode

Coding

IFBench

Instruction following

CritPt

Physics reasoning

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.

Artificial Analysis Intelligence Index v4.0 includes: GDPval-AA, 𝜏²-Bench Telecom, Terminal-Bench Hard, SciCode, AA-LCR, AA-Omniscience, IFBench, Humanity's Last Exam, GPQA Diamond, CritPt. See Intelligence Index methodology for further details, including a breakdown of each evaluation and how we run them.

Openness

Artificial Analysis Openness Index: Score

Openness Index assesses model openness on a 0 to 100 normalized scale (higher is more open)
Reasoning models are indicated by a lightbulb icon

Intelligence Index Comparisons

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.

Artificial Analysis Intelligence Index v4.0 includes: GDPval-AA, 𝜏²-Bench Telecom, Terminal-Bench Hard, SciCode, AA-LCR, AA-Omniscience, IFBench, Humanity's Last Exam, GPQA Diamond, CritPt. See Intelligence Index methodology for further details, including a breakdown of each evaluation and how we run them.

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). Blended price charts use Anthropic cache write price for the input leg.
  • 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).

Intelligence Index Token Use & Cost

Output Tokens Used to Run Artificial Analysis Intelligence Index

Tokens used to run all evaluations in the Artificial Analysis Intelligence Index
Reasoning models are indicated by a lightbulb icon

The number of tokens required to run all evaluations in the Artificial Analysis Intelligence Index (excluding repeats).

Cost to Run Artificial Analysis Intelligence Index

Cost (USD) to run all evaluations in the Artificial Analysis Intelligence Index
Reasoning models are indicated by a lightbulb icon

The cost to run the evaluations in the Artificial Analysis Intelligence Index, calculated using the model's input and output token pricing and the number of tokens used across evaluations (excluding repeats).

Context Window

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).

PricingUpdated

Pricing: Cache Hit, Input, and Output

Price (USD per M Tokens)
Reasoning models are indicated by a lightbulb icon

Price per token for cached prompts (previously processed), typically offering a significant discount compared to regular input price, represented as USD per million tokens. The values shown here are the cache hit price; cache write and cache storage are billed separately and vary by provider — see "Cache pricing by provider" for detail.

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

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). Blended price charts use Anthropic cache write price for the input leg.
  • 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.

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

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).

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).

Output Speed vs. Price

Output speed: output tokens per second · USD per 1M tokens (blended)
Most attractive quadrant
Alibaba
Anthropic
DeepSeek
Google
Kimi
Microsoft
MiniMax
Mistral
NVIDIA
OpenAI
xAI
Xiaomi
Z AI
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).

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

Seconds to output 500 tokens, including reasoning model 'thinking' time · Lower is better
Reasoning models are indicated by a lightbulb icon

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 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).

Model Size (Open Weights Models Only)

Model Size: Total and Active Parameters

Comparison between total model parameters and parameters active during inference
Reasoning models are indicated by a lightbulb icon

The total number of trainable weights and biases in the model, expressed in billions. These parameters are learned during training and determine the model's ability to process and generate responses.

The number of parameters actually executed during each inference forward pass, expressed in billions. For Mixture of Experts (MoE) models, a routing mechanism selects a subset of experts per token, resulting in fewer active than total parameters. Dense models use all parameters, so active equals total.

Frequently Asked Questions

Common questions about Phi-4

Phi-4 was released on December 12, 2024.

Phi-4 was created by Microsoft.

Phi-4 scores 10 on the Artificial Analysis Intelligence Index, placing it below average among other open weight non-reasoning models of similar size (median: 12).

Phi-4 generates output at 35.7 tokens per second (based on Microsoft's API), which is at the lower end compared to other open weight non-reasoning models of similar size (median: 95.3 t/s).

Phi-4 has a time to first token (TTFT) of 2.13s (based on Microsoft's API), which is somewhat higher than average compared to other open weight non-reasoning models of similar size (median: 1.63s).

Phi-4 costs $0.13 per 1M input tokens (better than average, median: $0.13) and $0.50 per 1M output tokens (somewhat higher than average, median: $0.25), based on Microsoft's API.

Phi-4 costs $0.13 per 1M input tokens and $0.50 per 1M output tokens (based on Microsoft's API). For a blended rate (7:2:1 cache hit/input/output ratio), this is $0.16 per 1M tokens. Pricing may vary by provider. Compare provider pricing

No, Phi-4 is not a reasoning model. It provides direct responses without extended chain-of-thought reasoning.

Phi-4 supports text only input.

Phi-4 supports text only output.

No, Phi-4 does not support image input. It can only process text.

No, Phi-4 is not multimodal. It only supports text only input.

Phi-4 has a context window of 16k tokens. This determines how much text and conversation history the model can process in a single request.

Yes, Phi-4 is open weights. The model weights are publicly available and can be downloaded for self-hosting.

Phi-4 has 14 billion parameters.

Phi-4 is released under the MIT license. This license allows commercial use. View license

Phi-4 achieves a score of 10 on the Artificial Analysis Intelligence Index. This composite benchmark evaluates models across reasoning, knowledge, mathematics, and coding.

Phi-4 has a knowledge cutoff of June 2024. The model's training data includes information up to this date.

Phi-4 is an open weights model that can be self-hosted. View providers

Phi-4 is an open weights model that can be downloaded and self-hosted. Compare providers