Skip to content
Get an API key

Live · serverless · 2 regions

Llama 4 Maverick

by Meta

A larger mixture-of-experts model that reads images and text and gives more careful answers than Scout.

  • chat
  • vision
  • FP8
Serverless priceUSD
Input
$0.27per 1M tokens
Output
$0.85per 1M tokens

Billed per token, no minimums.

Reference

Specifications

Model ID
llama-4-maverick
Endpoint
POST /v1/chat/completions
Publisher
Meta
Tasks
chat · vision
Served context
1M tokens
Quantization
FP8
Decode speed
118 tok/smedian, one request
p50 first token
165 mslast 24 hours
Regions
us-east · us-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

165 msmedian

p95 179 ms

Decode speed · tok/s

118 tok/smedian

p5 111 tok/s

Show the data
Median of each 15-minute step over the last 24 hours, newest first
StepTime to first tokenDecode speed
15m ago168 ms117 tok/s
30m ago170 ms118 tok/s
45m ago171 ms116 tok/s
1h ago170 ms115 tok/s
1h 15m ago173 ms118 tok/s
1h 30m ago172 ms119 tok/s
1h 45m ago173 ms118 tok/s
2h ago170 ms117 tok/s
2h 15m ago172 ms113 tok/s
2h 30m ago175 ms111 tok/s
2h 45m ago164 ms120 tok/s
3h ago165 ms120 tok/s
3h 15m ago179 ms114 tok/s
3h 30m ago173 ms118 tok/s
3h 45m ago167 ms116 tok/s
4h ago178 ms117 tok/s
4h 15m ago165 ms119 tok/s
4h 30m ago173 ms115 tok/s
4h 45m ago176 ms114 tok/s
5h ago177 ms114 tok/s
5h 15m ago174 ms118 tok/s
5h 30m ago164 ms117 tok/s
5h 45m ago163 ms119 tok/s
6h ago241 ms90 tok/s
6h 15m ago164 ms119 tok/s
6h 30m ago162 ms117 tok/s
6h 45m ago163 ms118 tok/s
7h ago163 ms117 tok/s
7h 15m ago160 ms122 tok/s
7h 30m ago165 ms117 tok/s
7h 45m ago160 ms120 tok/s
8h ago175 ms114 tok/s
8h 15m ago171 ms117 tok/s
8h 30m ago167 ms118 tok/s
8h 45m ago168 ms117 tok/s
9h ago164 ms121 tok/s
9h 15m ago161 ms123 tok/s
9h 30m ago160 ms120 tok/s
9h 45m ago166 ms120 tok/s
10h ago160 ms117 tok/s
10h 15m ago169 ms117 tok/s
10h 30m ago161 ms120 tok/s
10h 45m ago164 ms119 tok/s
11h ago164 ms118 tok/s
11h 15m ago154 ms124 tok/s
11h 30m ago157 ms125 tok/s
11h 45m ago169 ms117 tok/s
12h ago168 ms120 tok/s
12h 15m ago164 ms117 tok/s
12h 30m ago163 ms116 tok/s
12h 45m ago154 ms125 tok/s
13h ago207 ms100 tok/s
13h 15m ago167 ms115 tok/s
13h 30m ago161 ms122 tok/s
13h 45m ago157 ms124 tok/s
14h ago167 ms115 tok/s
14h 15m ago157 ms124 tok/s
14h 30m ago156 ms123 tok/s
14h 45m ago152 ms125 tok/s
15h ago164 ms121 tok/s
15h 15m ago151 ms125 tok/s
15h 30m ago163 ms122 tok/s
15h 45m ago165 ms121 tok/s
16h ago157 ms118 tok/s
16h 15m ago157 ms122 tok/s
16h 30m ago153 ms121 tok/s
16h 45m ago154 ms120 tok/s
17h ago214 ms104 tok/s
17h 15m ago168 ms117 tok/s
17h 30m ago156 ms119 tok/s
17h 45m ago167 ms116 tok/s
18h ago162 ms120 tok/s
18h 15m ago156 ms119 tok/s
18h 30m ago168 ms114 tok/s
18h 45m ago163 ms117 tok/s
19h ago165 ms122 tok/s
19h 15m ago154 ms124 tok/s
19h 30m ago155 ms123 tok/s
19h 45m ago164 ms118 tok/s
20h ago157 ms121 tok/s
20h 15m ago165 ms120 tok/s
20h 30m ago171 ms117 tok/s
20h 45m ago159 ms121 tok/s
21h ago170 ms116 tok/s
21h 15m ago170 ms119 tok/s
21h 30m ago167 ms121 tok/s
21h 45m ago169 ms120 tok/s
22h ago172 ms118 tok/s
22h 15m ago169 ms115 tok/s
22h 30m ago167 ms116 tok/s
22h 45m ago175 ms113 tok/s
23h ago217 ms101 tok/s
23h 15m ago172 ms117 tok/s
23h 30m ago167 ms119 tok/s
23h 45m ago176 ms116 tok/s
24h ago165 ms121 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": "llama-4-maverick",
    "messages": [
      { "role": "user", "content": "What does a KV cache do?" }
    ]
  }'
POST /v1/chat/completionsapi.weir.example

Dedicated

Need reserved capacity?

Llama 4 Maverick 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