Qwen3.8-27B

This Flask endpoint returns 500 under concurrent load.
Find the race condition and patch it.

The counter increment is not atomic, so two requests can read the same
value before either writes back. Wrapping the update in a lock, or moving
the counter to an atomic backend, closes the lost-update window.
Patch follows.

[chart image] What does this throughput curve tell us?

Throughput peaks at concurrency 32 and declines beyond it, which means
the batch scheduler saturates before the memory bound is reached.

from openai import OpenAI

client = OpenAI(
    api_key="YOUR_API_KEY",
    base_url="https://ap-1.aieev.cloud/endpoints/qwen-3-8-27b/v1"
)

response = client.chat.completions.create(
    model="qwen/qwen3.8-27b",
    messages=[{"role": "user", "content": "Hello!"}]
)

print(response.choices[0].message.content)

curl https://ap-1.aieev.cloud/endpoints/qwen-3-8-27b/v1/chat/completions \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "qwen/qwen3.8-27b",
    "messages": [{"role": "user", "content": "Hello!"}]
  }'

import OpenAI from "openai";

const client = new OpenAI({
  apiKey: "YOUR_API_KEY",
  baseURL: "https://ap-1.aieev.cloud/endpoints/qwen-3-8-27b/v1",
});

const response = await client.chat.completions.create({
  model: "qwen/qwen3.8-27b",
  messages: [{ role: "user", content: "Hello!" }],
});

console.log(response.choices[0].message.content);