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
- Endpoint
- Publisher
- Tasks
- Served context
- Quantization
- Decode speed
- p50 first token
- Regions
- Dedicated
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
| Step | Time to first token | Decode speed |
|---|---|---|
| 15m ago | 90 ms | 199 tok/s |
| 30m ago | 92 ms | 189 tok/s |
| 45m ago | 87 ms | 202 tok/s |
| 1h ago | 114 ms | 162 tok/s |
| 1h 15m ago | 87 ms | 199 tok/s |
| 1h 30m ago | 87 ms | 196 tok/s |
| 1h 45m ago | 92 ms | 188 tok/s |
| 2h ago | 89 ms | 191 tok/s |
| 2h 15m ago | 90 ms | 192 tok/s |
| 2h 30m ago | 86 ms | 194 tok/s |
| 2h 45m ago | 90 ms | 196 tok/s |
| 3h ago | 89 ms | 190 tok/s |
| 3h 15m ago | 84 ms | 198 tok/s |
| 3h 30m ago | 89 ms | 193 tok/s |
| 3h 45m ago | 91 ms | 198 tok/s |
| 4h ago | 89 ms | 200 tok/s |
| 4h 15m ago | 88 ms | 196 tok/s |
| 4h 30m ago | 91 ms | 198 tok/s |
| 4h 45m ago | 90 ms | 189 tok/s |
| 5h ago | 86 ms | 199 tok/s |
| 5h 15m ago | 84 ms | 202 tok/s |
| 5h 30m ago | 90 ms | 198 tok/s |
| 5h 45m ago | 90 ms | 193 tok/s |
| 6h ago | 129 ms | 153 tok/s |
| 6h 15m ago | 83 ms | 198 tok/s |
| 6h 30m ago | 89 ms | 200 tok/s |
| 6h 45m ago | 89 ms | 189 tok/s |
| 7h ago | 86 ms | 203 tok/s |
| 7h 15m ago | 89 ms | 191 tok/s |
| 7h 30m ago | 86 ms | 195 tok/s |
| 7h 45m ago | 87 ms | 201 tok/s |
| 8h ago | 83 ms | 205 tok/s |
| 8h 15m ago | 87 ms | 198 tok/s |
| 8h 30m ago | 89 ms | 194 tok/s |
| 8h 45m ago | 85 ms | 196 tok/s |
| 9h ago | 84 ms | 205 tok/s |
| 9h 15m ago | 85 ms | 203 tok/s |
| 9h 30m ago | 85 ms | 196 tok/s |
| 9h 45m ago | 83 ms | 200 tok/s |
| 10h ago | 87 ms | 196 tok/s |
| 10h 15m ago | 81 ms | 203 tok/s |
| 10h 30m ago | 82 ms | 202 tok/s |
| 10h 45m ago | 85 ms | 204 tok/s |
| 11h ago | 80 ms | 205 tok/s |
| 11h 15m ago | 86 ms | 196 tok/s |
| 11h 30m ago | 85 ms | 195 tok/s |
| 11h 45m ago | 79 ms | 201 tok/s |
| 12h ago | 79 ms | 207 tok/s |
| 12h 15m ago | 86 ms | 194 tok/s |
| 12h 30m ago | 80 ms | 204 tok/s |
| 12h 45m ago | 85 ms | 196 tok/s |
| 13h ago | 81 ms | 207 tok/s |
| 13h 15m ago | 85 ms | 198 tok/s |
| 13h 30m ago | 86 ms | 197 tok/s |
| 13h 45m ago | 81 ms | 206 tok/s |
| 14h ago | 79 ms | 203 tok/s |
| 14h 15m ago | 80 ms | 202 tok/s |
| 14h 30m ago | 83 ms | 200 tok/s |
| 14h 45m ago | 85 ms | 196 tok/s |
| 15h ago | 80 ms | 211 tok/s |
| 15h 15m ago | 81 ms | 204 tok/s |
| 15h 30m ago | 80 ms | 202 tok/s |
| 15h 45m ago | 83 ms | 208 tok/s |
| 16h ago | 78 ms | 210 tok/s |
| 16h 15m ago | 80 ms | 211 tok/s |
| 16h 30m ago | 83 ms | 201 tok/s |
| 16h 45m ago | 80 ms | 212 tok/s |
| 17h ago | 80 ms | 205 tok/s |
| 17h 15m ago | 81 ms | 211 tok/s |
| 17h 30m ago | 80 ms | 208 tok/s |
| 17h 45m ago | 79 ms | 211 tok/s |
| 18h ago | 86 ms | 197 tok/s |
| 18h 15m ago | 79 ms | 213 tok/s |
| 18h 30m ago | 86 ms | 200 tok/s |
| 18h 45m ago | 79 ms | 202 tok/s |
| 19h ago | 79 ms | 209 tok/s |
| 19h 15m ago | 80 ms | 210 tok/s |
| 19h 30m ago | 84 ms | 204 tok/s |
| 19h 45m ago | 86 ms | 200 tok/s |
| 20h ago | 85 ms | 205 tok/s |
| 20h 15m ago | 81 ms | 203 tok/s |
| 20h 30m ago | 83 ms | 204 tok/s |
| 20h 45m ago | 85 ms | 194 tok/s |
| 21h ago | 83 ms | 203 tok/s |
| 21h 15m ago | 82 ms | 198 tok/s |
| 21h 30m ago | 84 ms | 204 tok/s |
| 21h 45m ago | 85 ms | 195 tok/s |
| 22h ago | 86 ms | 201 tok/s |
| 22h 15m ago | 84 ms | 202 tok/s |
| 22h 30m ago | 87 ms | 194 tok/s |
| 22h 45m ago | 83 ms | 206 tok/s |
| 23h ago | 87 ms | 193 tok/s |
| 23h 15m ago | 88 ms | 203 tok/s |
| 23h 30m ago | 85 ms | 201 tok/s |
| 23h 45m ago | 88 ms | 196 tok/s |
| 24h ago | 84 ms | 201 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?" }
]
}'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 engineerCompare