Start free trial

Microsoft: Phi 4

microsoft/phi-4
CompareAPITry this model

[Microsoft Research](/microsoft) Phi-4 is designed to perform well in complex reasoning tasks and can operate efficiently in situations with limited memory or where quick responses are needed.

Modalities
In / out per 1M$0.07 / $0.14
Context16K
ReleasedJan 10, 2025

Providers

2 providers serve this model upstream.

Providers serving this model
ProviderInput /MOutput /MCache read /MLatency P50Throughput P50Uptime
DeepInfraus · bf16$0.06$0.11Variable1.45s40 tps100%
DeepInfraus · bf16$0.07$0.14Variable0.44s77 tps99.92%

Market-wide measurements, refreshed every 15 minutes. They describe who serves Phi 4 in general, not how ProjectCOZY routes your requests.

Pricing

List rates per 1M tokens — what Phi 4 costs pay-as-you-go elsewhere, not what you pay here.

Input
$0.07
per 1M tokens
Output
$0.14
per 1M tokens
Cache read
—
not published
Included in every pass
Phi 4 is included in every ProjectCOZY pass at no extra cost — one flat rate, no per-token metering. The rates above are there for comparison.

Performance

Time to first token and output speed across every provider, over the last 3 days.

0.0s1.0s2.0s3.0s4.0s
Oct 29, 12 AMOct 29, 8 AMOct 29, 4 PMOct 30, 12 AMOct 30, 8 AMOct 30, 4 PMOct 31, 12 AMOct 31, 8 AMOct 31, 4 PMNov 1, 12 AMNov 1, 8 AM
DeepInfra (us) DeepInfra (us)
0 tok/s50 tok/s100 tok/s150 tok/s200 tok/s
Oct 29, 12 AMOct 29, 8 AMOct 29, 4 PMOct 30, 12 AMOct 30, 8 AMOct 30, 4 PMOct 31, 12 AMOct 31, 8 AMOct 31, 4 PMNov 1, 12 AMNov 1, 8 AM
DeepInfra (us) DeepInfra (us)

Uptime

100% of requests were served successfully over the last 3 days, across all providers.

Last 3 daysAvailability 99.97%
Oct 29, 6:00 PMNow
Availability over the last 24 hoursAvailability 99.86%
100% 95% 90% 85%
Oct 31, 6:10 PMNow

Apps

Tools that most often run Phi 4, by tokens this week.

Concept Task Evaluation91.6% of tokens4.1BNAACLnoise Exp2 AlltoAlgebra | Llama | Algebra | 10-shot | prm5.6% of tokens250.5Mspoked-review1% of tokens46.4Mcrowd_crossover1% of tokens45.0MSpin0.8% of tokens34.7M

Activity

7.3B tokens processed market-wide in the last month.

0500.0M1.0B1.5B2.0B
Oct 2Oct 7Oct 12Oct 17Oct 22Oct 27Nov 1
Tokens

FAQ

Which model id do I use for Phi 4?
microsoft/phi-4. Send it as the model field from any OpenAI-compatible client, against https://minuykh.app/v1.
How much context does Phi 4 hold?
16,384 tokens per request, of which up to 14,745 can be the response.
What can it accept and return?
Input: text. Output: text.
What does Phi 4 cost on a pass?
Nothing beyond the pass itself. Pay-as-you-go elsewhere it lists at $0.07 per 1M input tokens and $0.14 per 1M output tokens.
How current is its knowledge?
Training data runs to Jun 2024. Anything later has to come from your prompt or a tool call.
Who built Phi 4?
Microsoft, in the Phi family.

Explore

Other models your pass can call.

Aion 3.5aion-labs/aion-3.5256KAionAion 3.5 Miniaion-labs/aion-3.5-mini256KAionAion-2.0aion-labs/aion-2.0128KAionAion-3.0aion-labs/aion-3.0128KAionAion-3.0-Miniaion-labs/aion-3.0-mini128KAionAion-RP 1.0 (8B)aion-labs/aion-rp-llama-3.1-8b32KAion