{ code }
Highly specialized auxiliary coding models designed to make AI codegen products faster and more reliable.

Edit Snippet
Response
TTFT: 541 ms
TPS: 2504 tok/s








Models
Custom SOTA models for coding workflows
Relace models are trained in-house to outperform frontier LLMs on specific steps in autonomous codegen. Our retrieval and code merging models are already state-of-the-art, and they continue to improve with each training run.
Enqueued
Startup
Execution
Status
09:34:22 AM
0.0s
-
09:34:19 AM
0.0s
0.0s
09:34:15 AM
0.0s
2.2s
09:34:14 AM
0.0s
1.9s
09:34:12 AM
0.0s
1.7s
09:34:10 AM
0.0s
2.1s
Deployments
Optimized deployments for fast inference
Each model is hosted with a custom inference engine, tuned to achieve nearly instantaneous output. Our code merging model runs at >2500 tok/s, and our retrieval models can search an entire codebase in <1s.
Integration
A competitive edge starts with only a few extra lines of code
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
import requests
url = "https://instantapply.endpoint.relace.run/v1/code/apply"
headers = {
"Content-Type": "application/json",
"Authorization": "Bearer [YOUR_API_KEY]"
}
initial_code = """def hello_world():
# TODO: Implement Hello World
pass"""
edit_snippet = """// ... keep existing code
print("Hello, world!")
// ... keep existing code"""
data = {
"initialCode": initial_code,
"editSnippet": edit_snippet
}
from relace import OpenAI client = OpenAI( api_key= "OPENAI_API_KEY", relace_api_key= "RELACE_API_KEY", ) messages = [ { "role": "system", "content": "..." }, {"role": "user", "content": "..." } ] completion = client.chat.completions.create( model="gpt-4o", messages= messages, )
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
import requests
url = "https://ranker.endpoint.relace.run/v1/code/rank"
headers = { "Content-Type": "application/json", "Authorization": "Bearer [YOUR_API_KEY]" }
query = "Change hello world to goodbye world."
codebase = [
{ "filename": "greeting.py", "code": "print('hello world')" },
...
]
data = { "query": query, "codebase": codebase, "token_limit": 128000 }
response = requests.post(url, headers=headers, json=data)
ranked_code = response.json()
import OpenAI from 'relace'; const client = new OpenAI({ apiKey: "OPENAI_API_KEY", relaceApiKey: "RELACE_API_KEY" }); const messages = [ { role: "system", content: "..." }, { role: "user", content: "..." } ]; const completion = await client.chat.completions.create({ model: "gpt-4o", messages: messages });
OUR INVESTORS




Testimonials