DOCS · v2

API reference

devseek speaks the OpenAI Chat Completions protocol. Any OpenAI SDK works once you change the base URL.

BASE URL

https://devseek.space/v1

AUTH

Authorization: Bearer dsk-…

MODEL

deepseek-v4.1-flash-free

224K tokens · 64K tokens

Quick start

OPENAI SDK · UNCHANGED

1curl https://devseek.space/v1/chat/completions \
2  -H "Authorization: Bearer $DEVSEEK_API_KEY" \
3  -H "Content-Type: application/json" \
4  -d '{
5    "model": "deepseek-v4.1-flash-free",
6    "stream": true,
7    "messages": [{"role": "user", "content": "Write a binary search in Rust"}]
8  }'

1import os
2from openai import OpenAI
3 
4client = OpenAI(
5    api_key=os.environ["DEVSEEK_API_KEY"],
6    base_url="https://devseek.space/v1",
7)
8 
9stream = client.chat.completions.create(
10    model="deepseek-v4.1-flash-free",
11    messages=[{"role": "user", "content": "Write a binary search in Rust"}],
12    max_tokens=4096,
13    stream=True,
14)
15for chunk in stream:
16    if chunk.choices:
17        print(chunk.choices[0].delta.content or "", end="")

1import OpenAI from "openai";
2 
3const client = new OpenAI({
4  apiKey: process.env.DEVSEEK_API_KEY,
5  baseURL: "https://devseek.space/v1",
6});
7 
8const stream = await client.chat.completions.create({
9  model: "deepseek-v4.1-flash-free",
10  messages: [{ role: "user", content: "Review this diff for bugs" }],
11  stream: true,
12});
13for await (const c of stream) process.stdout.write(c.choices[0]?.delta?.content ?? "");

1from openai import OpenAI
2client = OpenAI(api_key=KEY, base_url="https://devseek.space/v1")
3 
4tools = [{"type": "function", "function": {
5    "name": "run_tests",
6    "description": "Run the test suite and return failures",
7    "parameters": {"type": "object",
8                   "properties": {"path": {"type": "string"}},
9                   "required": ["path"]}}}]
10 
11r = client.chat.completions.create(
12    model="deepseek-v4.1-flash-free", tools=tools,
13    messages=[{"role": "user", "content": "Tests in src/ fail. Investigate."}])
14print(r.choices[0].message.tool_calls)

1# opencode.json — add devseek as an OpenAI-compatible provider
2{
3  "provider": {
4    "devseek": {
5      "npm": "@ai-sdk/openai-compatible",
6      "options": { "baseURL": "https://devseek.space/v1", "apiKey": "{env:DEVSEEK_API_KEY}" },
7      "models": { "deepseek-v4.1-flash-free": { "limit": { "context": 224000, "output": 65536 } } }
8    }
9  }
10}
11 
12# aider
13export OPENAI_API_BASE=https://devseek.space/v1 OPENAI_API_KEY=$DEVSEEK_API_KEY
14aider --model openai/deepseek-v4.1-flash-free

OpenAI-compatible /v1/chat/completionsAPI reference

Endpoints

Route	What it does	Notes
`GET /v1/models`	List models	Needs a key
`POST /v1/chat/completions`	Chat, streaming or not, with tools	stream: true for SSE; a usage chunk comes last
`GET /api/status`	Public status JSON	No key
`GET /health`	Liveness	No key

Limits

Input is limited to 224K tokens and output to 64K tokens. We estimate input size from the request bytes (about 3.3 bytes per token) before calling the model, so a request near the limit may be refused a little early.

Input above the limit returns 400 with code context_length_exceeded. Nothing is sent to the model.
max_tokens above 64K tokens returns 400 with code invalid_max_tokens. devseek never lowers or rewrites your max_tokens.
Request bodies are limited to 4 MB.

Sessions and levels

A session is one chat completion request that has started and not yet finished. Your level sets how many sessions may run at once:

2L1–10Script
2–3L11–20Function
3–4L21–30Module
4–5L31–40Package
5L41–50Library
6L51–60Framework
6–7L61–70Runtime
7–8L71–80Compiler
8–9L81–90Kernel
9–10L91–100Architect

Over your limit you get:

HTTP/1.1 429 Too Many Requests
Retry-After: 1
{"error":{"message":"Level 1 allows 3 concurrent sessions; 3 are already running. …",
 "type":"rate_limit_error","code":"session_limit_reached",
 "level":1,"session_limit":3,"sessions_in_use":3}}

Agents and editors

Any tool that accepts an OpenAI-compatible provider works. Use the base URL, your key and the model ID below. Keep parallel tool workers at or below your session limit.

Errors

Status / code	Meaning	What to do
`401 missing_api_key / invalid_api_key`	Missing or wrong key	Send Authorization: Bearer dsk-…
`400 invalid_json / invalid_messages`	Malformed body or messages	Fix the request; do not retry as is
`400 context_length_exceeded`	Input over 224K tokens	Shorten or summarise the conversation
`404 model_not_found`	Unknown model	Use deepseek-v4.1-flash-free
`413 body_too_large`	Body over 4 MB	Send less context
`429 session_limit_reached`	Too many concurrent sessions	Wait for a request to finish, then retry
`503 upstream_saturated / upstream_busy`	devseek or the model is at capacity	Retry after the Retry-After header
`502 upstream_error`	The model backend failed	Safe to retry