Skip to content
Get an API key

Live · serverless · 3 regions

Devstral Small 2

by Mistral AI

A 24B model tuned for software engineering agents that open files, run commands and fix failing tests.

  • code
  • FP8
Serverless priceUSD
Input
$0.09per 1M tokens
Output
$0.30per 1M tokens

Billed per token, no minimums.

Reference

Specifications

Model ID
devstral-small-2
Endpoint
POST /v1/chat/completions
Publisher
Mistral AI
Tasks
code
Served context
128K tokens
Quantization
FP8
Decode speed
160 tok/smedian, one request
p50 first token
120 mslast 24 hours
Regions
us-east · us-west · 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

120 msmedian

p95 148 ms

Decode speed · tok/s

160 tok/smedian

p5 143 tok/s

Show the data
Median of each 15-minute step over the last 24 hours, newest first
StepTime to first tokenDecode speed
15m ago117 ms165 tok/s
30m ago124 ms158 tok/s
45m ago126 ms159 tok/s
1h ago128 ms153 tok/s
1h 15m ago121 ms160 tok/s
1h 30m ago184 ms120 tok/s
1h 45m ago126 ms159 tok/s
2h ago122 ms163 tok/s
2h 15m ago126 ms153 tok/s
2h 30m ago124 ms156 tok/s
2h 45m ago119 ms164 tok/s
3h ago124 ms153 tok/s
3h 15m ago130 ms153 tok/s
3h 30m ago122 ms156 tok/s
3h 45m ago121 ms158 tok/s
4h ago126 ms154 tok/s
4h 15m ago118 ms158 tok/s
4h 30m ago123 ms159 tok/s
4h 45m ago122 ms155 tok/s
5h ago173 ms120 tok/s
5h 15m ago120 ms163 tok/s
5h 30m ago127 ms154 tok/s
5h 45m ago123 ms160 tok/s
6h ago121 ms157 tok/s
6h 15m ago122 ms160 tok/s
6h 30m ago126 ms151 tok/s
6h 45m ago124 ms160 tok/s
7h ago122 ms155 tok/s
7h 15m ago126 ms159 tok/s
7h 30m ago117 ms163 tok/s
7h 45m ago117 ms162 tok/s
8h ago122 ms159 tok/s
8h 15m ago118 ms163 tok/s
8h 30m ago124 ms156 tok/s
8h 45m ago122 ms158 tok/s
9h ago119 ms159 tok/s
9h 15m ago166 ms125 tok/s
9h 30m ago118 ms158 tok/s
9h 45m ago118 ms158 tok/s
10h ago124 ms155 tok/s
10h 15m ago121 ms160 tok/s
10h 30m ago113 ms165 tok/s
10h 45m ago148 ms143 tok/s
11h ago121 ms163 tok/s
11h 15m ago122 ms163 tok/s
11h 30m ago112 ms164 tok/s
11h 45m ago115 ms164 tok/s
12h ago113 ms166 tok/s
12h 15m ago110 ms163 tok/s
12h 30m ago116 ms166 tok/s
12h 45m ago121 ms160 tok/s
13h ago119 ms164 tok/s
13h 15m ago111 ms166 tok/s
13h 30m ago111 ms167 tok/s
13h 45m ago112 ms164 tok/s
14h ago109 ms168 tok/s
14h 15m ago110 ms164 tok/s
14h 30m ago118 ms163 tok/s
14h 45m ago116 ms165 tok/s
15h ago120 ms159 tok/s
15h 15m ago109 ms169 tok/s
15h 30m ago147 ms145 tok/s
15h 45m ago115 ms160 tok/s
16h ago117 ms158 tok/s
16h 15m ago113 ms165 tok/s
16h 30m ago110 ms162 tok/s
16h 45m ago119 ms161 tok/s
17h ago112 ms162 tok/s
17h 15m ago120 ms161 tok/s
17h 30m ago117 ms160 tok/s
17h 45m ago116 ms161 tok/s
18h ago121 ms162 tok/s
18h 15m ago113 ms168 tok/s
18h 30m ago120 ms164 tok/s
18h 45m ago121 ms163 tok/s
19h ago119 ms163 tok/s
19h 15m ago122 ms154 tok/s
19h 30m ago122 ms160 tok/s
19h 45m ago114 ms167 tok/s
20h ago124 ms161 tok/s
20h 15m ago113 ms164 tok/s
20h 30m ago113 ms168 tok/s
20h 45m ago116 ms159 tok/s
21h ago121 ms163 tok/s
21h 15m ago120 ms160 tok/s
21h 30m ago117 ms163 tok/s
21h 45m ago115 ms167 tok/s
22h ago157 ms137 tok/s
22h 15m ago118 ms161 tok/s
22h 30m ago124 ms157 tok/s
22h 45m ago120 ms155 tok/s
23h ago119 ms163 tok/s
23h 15m ago121 ms159 tok/s
23h 30m ago117 ms158 tok/s
23h 45m ago126 ms155 tok/s
24h ago127 ms155 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": "devstral-small-2",
    "messages": [
      { "role": "user", "content": "What does a KV cache do?" }
    ]
  }'
POST /v1/chat/completionsapi.weir.example

Dedicated

Need reserved capacity?

Devstral Small 2 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