Skip to content
Get an API key

Live · serverless · 3 regions

Mistral Small 3.2 24B

by Mistral AI

A 24B model that follows instructions closely, calls functions reliably and reads images, at a low price.

  • chat
  • vision
  • BF16
  • dedicated
Serverless priceUSD
Input
$0.10per 1M tokens
Output
$0.25per 1M tokens

Billed per token, no minimums.

Reference

Specifications

Model ID
mistral-small-3-2
Endpoint
POST /v1/chat/completions
Publisher
Mistral AI
Tasks
chat · vision
Served context
128K tokens
Quantization
BF16
Decode speed
170 tok/smedian, one request
p50 first token
110 mslast 24 hours
Regions
us-east · us-west · eu-west
Dedicated
1× NVIDIA H100 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

110 msmedian

p95 119 ms

Decode speed · tok/s

170 tok/smedian

p5 162 tok/s

Show the data
Median of each 15-minute step over the last 24 hours, newest first
StepTime to first tokenDecode speed
15m ago113 ms172 tok/s
30m ago115 ms170 tok/s
45m ago118 ms165 tok/s
1h ago109 ms167 tok/s
1h 15m ago114 ms169 tok/s
1h 30m ago109 ms174 tok/s
1h 45m ago116 ms165 tok/s
2h ago109 ms173 tok/s
2h 15m ago119 ms163 tok/s
2h 30m ago117 ms169 tok/s
2h 45m ago112 ms169 tok/s
3h ago113 ms165 tok/s
3h 15m ago116 ms166 tok/s
3h 30m ago119 ms168 tok/s
3h 45m ago156 ms133 tok/s
4h ago115 ms170 tok/s
4h 15m ago114 ms168 tok/s
4h 30m ago114 ms163 tok/s
4h 45m ago112 ms174 tok/s
5h ago114 ms167 tok/s
5h 15m ago112 ms168 tok/s
5h 30m ago108 ms169 tok/s
5h 45m ago111 ms170 tok/s
6h ago114 ms166 tok/s
6h 15m ago117 ms168 tok/s
6h 30m ago112 ms165 tok/s
6h 45m ago110 ms167 tok/s
7h ago108 ms174 tok/s
7h 15m ago108 ms176 tok/s
7h 30m ago111 ms165 tok/s
7h 45m ago110 ms170 tok/s
8h ago110 ms172 tok/s
8h 15m ago108 ms174 tok/s
8h 30m ago106 ms169 tok/s
8h 45m ago110 ms172 tok/s
9h ago143 ms147 tok/s
9h 15m ago111 ms169 tok/s
9h 30m ago114 ms164 tok/s
9h 45m ago105 ms171 tok/s
10h ago111 ms168 tok/s
10h 15m ago112 ms169 tok/s
10h 30m ago156 ms132 tok/s
10h 45m ago110 ms170 tok/s
11h ago104 ms171 tok/s
11h 15m ago104 ms177 tok/s
11h 30m ago102 ms174 tok/s
11h 45m ago102 ms175 tok/s
12h ago105 ms176 tok/s
12h 15m ago111 ms169 tok/s
12h 30m ago102 ms172 tok/s
12h 45m ago106 ms178 tok/s
13h ago107 ms173 tok/s
13h 15m ago139 ms151 tok/s
13h 30m ago101 ms174 tok/s
13h 45m ago105 ms178 tok/s
14h ago100 ms177 tok/s
14h 15m ago110 ms172 tok/s
14h 30m ago100 ms177 tok/s
14h 45m ago104 ms175 tok/s
15h ago107 ms169 tok/s
15h 15m ago105 ms172 tok/s
15h 30m ago108 ms174 tok/s
15h 45m ago101 ms174 tok/s
16h ago109 ms170 tok/s
16h 15m ago107 ms168 tok/s
16h 30m ago104 ms180 tok/s
16h 45m ago103 ms175 tok/s
17h ago107 ms171 tok/s
17h 15m ago104 ms175 tok/s
17h 30m ago109 ms169 tok/s
17h 45m ago101 ms180 tok/s
18h ago105 ms176 tok/s
18h 15m ago112 ms166 tok/s
18h 30m ago103 ms173 tok/s
18h 45m ago112 ms173 tok/s
19h ago103 ms178 tok/s
19h 15m ago113 ms166 tok/s
19h 30m ago112 ms170 tok/s
19h 45m ago107 ms175 tok/s
20h ago106 ms176 tok/s
20h 15m ago111 ms173 tok/s
20h 30m ago114 ms164 tok/s
20h 45m ago109 ms176 tok/s
21h ago114 ms170 tok/s
21h 15m ago107 ms176 tok/s
21h 30m ago113 ms167 tok/s
21h 45m ago107 ms170 tok/s
22h ago112 ms168 tok/s
22h 15m ago106 ms169 tok/s
22h 30m ago116 ms163 tok/s
22h 45m ago110 ms169 tok/s
23h ago115 ms162 tok/s
23h 15m ago108 ms168 tok/s
23h 30m ago117 ms165 tok/s
23h 45m ago108 ms172 tok/s
24h ago115 ms163 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": "mistral-small-3-2",
    "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× H100
Replica rate
$2.49 / hour
Peak throughput
9,000 tok/s
Break-even, 1 replica
10.4B 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