Skip to content
Get an API key

Live · serverless · 3 regions

Llama 4 Scout

by Meta

A mixture-of-experts model with a very long context window, useful for reading whole codebases or contract bundles in one request.

  • chat
  • vision
  • FP8
  • dedicated
Serverless priceUSD
Input
$0.18per 1M tokens
Output
$0.59per 1M tokens

Billed per token, no minimums.

Reference

Specifications

Model ID
llama-4-scout
Endpoint
POST /v1/chat/completions
Publisher
Meta
Tasks
chat · vision
Served context
1M tokens
Quantization
FP8
Decode speed
150 tok/smedian, one request
p50 first token
130 mslast 24 hours
Regions
us-east · us-west · eu-west
Dedicated
1× NVIDIA H200 per replica

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

130 msmedian

p95 176 ms

Decode speed · tok/s

150 tok/smedian

p5 124 tok/s

Show the data
Median of each 15-minute step over the last 24 hours, newest first
StepTime to first tokenDecode speed
15m ago130 ms149 tok/s
30m ago134 ms148 tok/s
45m ago134 ms145 tok/s
1h ago183 ms116 tok/s
1h 15m ago131 ms152 tok/s
1h 30m ago137 ms149 tok/s
1h 45m ago140 ms140 tok/s
2h ago133 ms149 tok/s
2h 15m ago177 ms119 tok/s
2h 30m ago137 ms146 tok/s
2h 45m ago131 ms150 tok/s
3h ago131 ms153 tok/s
3h 15m ago139 ms140 tok/s
3h 30m ago132 ms150 tok/s
3h 45m ago138 ms144 tok/s
4h ago129 ms146 tok/s
4h 15m ago137 ms149 tok/s
4h 30m ago129 ms153 tok/s
4h 45m ago134 ms150 tok/s
5h ago133 ms145 tok/s
5h 15m ago130 ms148 tok/s
5h 30m ago127 ms155 tok/s
5h 45m ago136 ms149 tok/s
6h ago130 ms146 tok/s
6h 15m ago180 ms124 tok/s
6h 30m ago137 ms149 tok/s
6h 45m ago135 ms147 tok/s
7h ago126 ms153 tok/s
7h 15m ago126 ms149 tok/s
7h 30m ago161 ms135 tok/s
7h 45m ago135 ms145 tok/s
8h ago133 ms149 tok/s
8h 15m ago131 ms149 tok/s
8h 30m ago135 ms146 tok/s
8h 45m ago122 ms155 tok/s
9h ago130 ms153 tok/s
9h 15m ago132 ms149 tok/s
9h 30m ago133 ms150 tok/s
9h 45m ago133 ms150 tok/s
10h ago124 ms158 tok/s
10h 15m ago123 ms152 tok/s
10h 30m ago155 ms140 tok/s
10h 45m ago130 ms146 tok/s
11h ago128 ms152 tok/s
11h 15m ago132 ms151 tok/s
11h 30m ago171 ms129 tok/s
11h 45m ago130 ms150 tok/s
12h ago129 ms153 tok/s
12h 15m ago118 ms161 tok/s
12h 30m ago129 ms155 tok/s
12h 45m ago120 ms156 tok/s
13h ago119 ms161 tok/s
13h 15m ago129 ms155 tok/s
13h 30m ago129 ms147 tok/s
13h 45m ago123 ms150 tok/s
14h ago128 ms151 tok/s
14h 15m ago129 ms148 tok/s
14h 30m ago127 ms149 tok/s
14h 45m ago118 ms156 tok/s
15h ago127 ms151 tok/s
15h 15m ago127 ms152 tok/s
15h 30m ago125 ms155 tok/s
15h 45m ago124 ms150 tok/s
16h ago123 ms155 tok/s
16h 15m ago123 ms151 tok/s
16h 30m ago163 ms134 tok/s
16h 45m ago125 ms150 tok/s
17h ago128 ms151 tok/s
17h 15m ago128 ms150 tok/s
17h 30m ago128 ms153 tok/s
17h 45m ago127 ms149 tok/s
18h ago123 ms151 tok/s
18h 15m ago128 ms153 tok/s
18h 30m ago121 ms156 tok/s
18h 45m ago132 ms149 tok/s
19h ago129 ms153 tok/s
19h 15m ago176 ms125 tok/s
19h 30m ago120 ms154 tok/s
19h 45m ago131 ms153 tok/s
20h ago123 ms155 tok/s
20h 15m ago176 ms124 tok/s
20h 30m ago125 ms150 tok/s
20h 45m ago129 ms154 tok/s
21h ago125 ms150 tok/s
21h 15m ago183 ms117 tok/s
21h 30m ago129 ms148 tok/s
21h 45m ago131 ms145 tok/s
22h ago133 ms151 tok/s
22h 15m ago126 ms156 tok/s
22h 30m ago124 ms157 tok/s
22h 45m ago125 ms157 tok/s
23h ago133 ms146 tok/s
23h 15m ago132 ms150 tok/s
23h 30m ago135 ms144 tok/s
23h 45m ago131 ms151 tok/s
24h ago132 ms151 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-scout",
    "messages": [
      { "role": "user", "content": "What does a KV cache do?" }
    ]
  }'
POST /v1/chat/completionsapi.weir.example

Dedicated

Run it on GPUs reserved for you

Replica
1× H200
Replica rate
$3.39 / hour
Peak throughput
9,000 tok/s
Break-even, 1 replica
6.43B tok/mo

Replicas reserved for you, billed by the second, with a 99.95% monthly uptime SLA. Break-even is where one replica costs the same as serverless, at half input and half output tokens.

Run on dedicated GPUsCompare costs

Compare

Similar models