Skip to content
Get an API key

Live · serverless · 2 regions

Phi-4 14B

by Microsoft

A compact 14B model trained on curated data, good at math and short reasoning chains for its size.

  • chat
  • reasoning
  • BF16
Serverless priceUSD
Input
$0.07per 1M tokens
Output
$0.14per 1M tokens

Billed per token, no minimums.

Reference

Specifications

Model ID
phi-4
Endpoint
POST /v1/chat/completions
Publisher
Microsoft
Tasks
chat · reasoning
Served context
16K tokens
Quantization
BF16
Decode speed
200 tok/smedian, one request
p50 first token
85 mslast 24 hours
Regions
us-east · eu-west
Dedicated
On request

Last 24 hours

Latency and speed

Median of each 15-minute step on the shared pool, measured at the API edge. Time to first token includes queueing; decode speed is for one request.

Time to first token · ms

85 msmedian

p95 91 ms

Decode speed · tok/s

200 tok/smedian

p5 189 tok/s

Show the data
Median of each 15-minute step over the last 24 hours, newest first
StepTime to first tokenDecode speed
15m ago90 ms199 tok/s
30m ago92 ms189 tok/s
45m ago87 ms202 tok/s
1h ago114 ms162 tok/s
1h 15m ago87 ms199 tok/s
1h 30m ago87 ms196 tok/s
1h 45m ago92 ms188 tok/s
2h ago89 ms191 tok/s
2h 15m ago90 ms192 tok/s
2h 30m ago86 ms194 tok/s
2h 45m ago90 ms196 tok/s
3h ago89 ms190 tok/s
3h 15m ago84 ms198 tok/s
3h 30m ago89 ms193 tok/s
3h 45m ago91 ms198 tok/s
4h ago89 ms200 tok/s
4h 15m ago88 ms196 tok/s
4h 30m ago91 ms198 tok/s
4h 45m ago90 ms189 tok/s
5h ago86 ms199 tok/s
5h 15m ago84 ms202 tok/s
5h 30m ago90 ms198 tok/s
5h 45m ago90 ms193 tok/s
6h ago129 ms153 tok/s
6h 15m ago83 ms198 tok/s
6h 30m ago89 ms200 tok/s
6h 45m ago89 ms189 tok/s
7h ago86 ms203 tok/s
7h 15m ago89 ms191 tok/s
7h 30m ago86 ms195 tok/s
7h 45m ago87 ms201 tok/s
8h ago83 ms205 tok/s
8h 15m ago87 ms198 tok/s
8h 30m ago89 ms194 tok/s
8h 45m ago85 ms196 tok/s
9h ago84 ms205 tok/s
9h 15m ago85 ms203 tok/s
9h 30m ago85 ms196 tok/s
9h 45m ago83 ms200 tok/s
10h ago87 ms196 tok/s
10h 15m ago81 ms203 tok/s
10h 30m ago82 ms202 tok/s
10h 45m ago85 ms204 tok/s
11h ago80 ms205 tok/s
11h 15m ago86 ms196 tok/s
11h 30m ago85 ms195 tok/s
11h 45m ago79 ms201 tok/s
12h ago79 ms207 tok/s
12h 15m ago86 ms194 tok/s
12h 30m ago80 ms204 tok/s
12h 45m ago85 ms196 tok/s
13h ago81 ms207 tok/s
13h 15m ago85 ms198 tok/s
13h 30m ago86 ms197 tok/s
13h 45m ago81 ms206 tok/s
14h ago79 ms203 tok/s
14h 15m ago80 ms202 tok/s
14h 30m ago83 ms200 tok/s
14h 45m ago85 ms196 tok/s
15h ago80 ms211 tok/s
15h 15m ago81 ms204 tok/s
15h 30m ago80 ms202 tok/s
15h 45m ago83 ms208 tok/s
16h ago78 ms210 tok/s
16h 15m ago80 ms211 tok/s
16h 30m ago83 ms201 tok/s
16h 45m ago80 ms212 tok/s
17h ago80 ms205 tok/s
17h 15m ago81 ms211 tok/s
17h 30m ago80 ms208 tok/s
17h 45m ago79 ms211 tok/s
18h ago86 ms197 tok/s
18h 15m ago79 ms213 tok/s
18h 30m ago86 ms200 tok/s
18h 45m ago79 ms202 tok/s
19h ago79 ms209 tok/s
19h 15m ago80 ms210 tok/s
19h 30m ago84 ms204 tok/s
19h 45m ago86 ms200 tok/s
20h ago85 ms205 tok/s
20h 15m ago81 ms203 tok/s
20h 30m ago83 ms204 tok/s
20h 45m ago85 ms194 tok/s
21h ago83 ms203 tok/s
21h 15m ago82 ms198 tok/s
21h 30m ago84 ms204 tok/s
21h 45m ago85 ms195 tok/s
22h ago86 ms201 tok/s
22h 15m ago84 ms202 tok/s
22h 30m ago87 ms194 tok/s
22h 45m ago83 ms206 tok/s
23h ago87 ms193 tok/s
23h 15m ago88 ms203 tok/s
23h 30m ago85 ms201 tok/s
23h 45m ago88 ms196 tok/s
24h ago84 ms201 tok/s

API

Make a request

The same request shape as every other model on /v1/chat/completions. Switching between them is a change to one string.

curl https://api.weir.example/v1/chat/completions \
  -H "Authorization: Bearer $WEIR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "phi-4",
    "messages": [
      { "role": "user", "content": "What does a KV cache do?" }
    ]
  }'
POST /v1/chat/completionsapi.weir.example

Dedicated

Need reserved capacity?

Phi-4 14B runs on the shared serverless pool. Dedicated replicas for it are set up on request, with the same API and the same model ID.

Talk to an engineer

Compare

Similar models