↑
How you go up
Each request is judged once: is it about software? Only then is XP added: 1 XP per 750 tokens (prompt plus completion, as counted by the model), up to 90,000 XP a day. When XP reaches the next level, you move up at the daily update.
FREE INFERENCE · FOR CODING
devseek is an OpenAI-compatible endpoint for deepseek-v4.1-flash-free, meant for coding. No card, no token meter. Your limit is how many requests run at once: it grows when you use it for code and shrinks when you use it for anything else.
$ curl https://devseek.space/v1/chat/completions \ -H "Authorization: Bearer $DEVSEEK_API_KEY" \ -d '{"model":"deepseek-v4.1-flash-free","stream":true, "messages":[{"role":"user", "content":"Fix the off-by-one in bsearch()"}]}'
Context window
224K tokensMax output
64K tokensStreaming
Server-sent eventsTool calls
OpenAI function callingAPI shape
OpenAI Chat CompletionsPrice
$0 · paid with dataHOW ACCESS WORKS
No credit card and no token budget. The only limit is how many requests you run at the same time.
Get API keyA Cloudflare Turnstile challenge, verified on our server, not only in your browser.
No GitHub check. Each request is classified as coding or not: local rules decide clear cases, jev decides unclear ones. Coding raises your level; other use lowers it.
The free service runs on data: your prompts and responses are stored and may be licensed (sold) in de-identified form. Without this agreement there is no key.
Same OpenAI client, new base URL. Streaming and tool calls included.
THE MODEL
devseek serves exactly one model so the numbers below mean something. They come from the requests this gateway handled in the last 24 hours.
Model ID
deepseek-v4.1-flash-free
Up to 224K tokens of context and 64K output. The upstream accepted 239,761 input tokens in our test; 224K leaves a safety margin.
—
Time to first token p50 · p95
—Full response p50 · p95
—Output speed
—Success, 24h
—Mock upstream: these numbers are a test double, not the real model.
LEVELS
Every key starts at level 1 with 2 concurrent sessions. The more tokens you spend on coding, the higher you climb, slowly, up to 10 sessions at level 100.
↑
Each request is judged once: is it about software? Only then is XP added: 1 XP per 750 tokens (prompt plus completion, as counted by the model), up to 90,000 XP a day. When XP reaches the next level, you move up at the daily update.
↓
A request that is not about code removes 1 XP per 1,500 tokens, up to 45,000 XP a day. A day without coding shrinks XP by 3%. You drop a level only when XP falls clearly below your level's threshold (85% of it).
±1
Levels change once a day, by at most one step up or down. A single burst of requests cannot jump you several levels.
STATUS · LIVE
State
Loading live data…
The status service did not answer. Try again in a minute.
Sessions in flight
—Requests, last hour
—Success, 24h
—Upstream capacity in use
—MADE FOR CODING
Use the Chat Completions format you already know. Keep your retries, your agent framework and your editor plugin.
import os
from openai import OpenAI
client = OpenAI(
api_key=os.environ["DEVSEEK_API_KEY"],
base_url="https://devseek.space/v1",
)
stream = client.chat.completions.create(
model="deepseek-v4.1-flash-free",
messages=[{"role": "user", "content": "Write a binary search in Rust"}],
max_tokens=4096,
stream=True,
)
for chunk in stream:
if chunk.choices:
print(chunk.choices[0].delta.content or "", end="")import OpenAI from "openai";
const client = new OpenAI({
apiKey: process.env.DEVSEEK_API_KEY,
baseURL: "https://devseek.space/v1",
});
const stream = await client.chat.completions.create({
model: "deepseek-v4.1-flash-free",
messages: [{ role: "user", content: "Review this diff for bugs" }],
stream: true,
});
for await (const c of stream) process.stdout.write(c.choices[0]?.delta?.content ?? "");curl https://devseek.space/v1/chat/completions \
-H "Authorization: Bearer $DEVSEEK_API_KEY" \
-H "Content-Type: application/json" \
-d '{
"model": "deepseek-v4.1-flash-free",
"stream": true,
"messages": [{"role": "user", "content": "Write a binary search in Rust"}]
}'
THE DEAL, IN PLAIN WORDS
devseek stores your prompts and the model's responses, keeps the raw records encrypted for up to 30 days, removes obvious personal data and secrets, and may share or license (sell) de-identified datasets to third parties such as AI labs and dataset brokers. This is the condition of the free service: there is no key without it. You can download or delete your stored data at any time.
Read the data use pageFree
$0No card. No token meter. No trial that ends.
Your data
Stored, de-identified and sold as datasets. Agreeing is required.
Retention
30 daysRaw records are deleted after 30 days; only de-identified data remains.
Two minutes from now you can have a key.