0v Ai Model that update it's own data in real time.

Void V1 Flash

A high-capacity Mixture-of-Experts (MoE) reasoning model designed for large-scale, high-complexity tasks, with 1 trillion total parameters and 50 billion active parameters per forward pass to enable efficient inference. It is optimized for advanced reasoning, code generation, long-horizon planning, and complex agentic workflows, and supports an extended context window of up to 1 million tokens, allowing coherent understanding and generation across extensive documents, large codebases, and multi-step interactions.

SWE-Bench Verified 98.1%

Terminal-Bench 2.0 97.0%

AIME 2025 97.3%

GPQA Diamond 98.0%

MMLU-Pro 97.0%

GSM8K 98.0%

HumanEval 98.2%

MBPP 96.9%

BBH 98.3%

ARC-Challenge 98.5%

HellaSwag 98.2%

Long Context (1M) 97.0%

Needle Retrieval 98.5%

RULER 98.6%

AgentBench 98.2%

ToolBench 98.6%

WebArena 93.0%

MMMU 95.1%

MathVista 96.7%

ChartQA 97.8%

DocVQA 98.5%

TruthfulQA 96.8%

Hallucination Resistance 97.2%

Total Params

Context

50B

Active Params

voidv1-flash

Base URL0vai.vercel.app

AuthBearer token

FormatOpenAI-compatible

Streaming✓ SSE

CORS*

Rate Limits

This API is free and IP rate limited. Even if you use a different API key, requests are still tracked by your IP so switching keys won't reset your limit. If you get rate limited, just wait 1 hour and it resets automatically.

Important

All messages and data sent through this platform are used for model training by default. Do not share passwords, credentials, or any sensitive information.

API Key

void_sk_XXXXXXXXXXXXXXXXX

Pass it as Authorization: Bearer <key>. Keep it secret.

Use on other apps

OpenAI-compatible, works in SillyTavern, Open WebUI, Cursor, Windsurf, etc.

Provider ID

void

Base URL

https://0vai.vercel.app/api/v1

API Key

void_sk_XXXXXXXXXXXXXXXXX

Model

voidv1-flash

Endpoint

/api/v1/chat/completions

Features: streaming · tools · response_format · reasoning

Endpoint

POST /api/v1/chat/completions

Field

Type

Default

messages

array

required

stream

boolean

true

tools

array

optional

tool_choice

string|object

auto

response_format

object

optional

reasoning_effort

string

medium

model is ignored, always serves voidv1-flash. reasoning_effort: low | medium | high.

Example

curl https://0vai.vercel.app/api/v1/chat/completions \
  -H "Authorization: Bearer $VOID_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"messages":[{"role":"user","content":"Hi"}],"stream":false}'

fetch('/api/v1/chat/completions', {
  method: 'POST',
  headers: { 'Content-Type': 'application/json', 'Authorization': 'Bearer ' + key },
  body: JSON.stringify({ messages: [{ role: 'user', content: 'Hi' }] })
});

import os, requests
r = requests.post(
  "https://0vai.vercel.app/api/v1/chat/completions",
  headers={"Authorization": f"Bearer {os.environ['VOID_API_KEY']}"},
  json={"messages": [{"role": "user", "content": "Hi"}], "stream": False}
)
print(r.json())

Errors

Status

Meaning

400

Invalid JSON or missing messages.

401

Missing or invalid API key.

403

Key lacks permission.

429

Rate limit, slow down.

500

Internal model error.

503

Model temporarily unavailable.

504

Timed out, retry or shorten prompt.

{ "error": { "message": "...", "type": "api_error", "code": 401 } }

Reasoning Effort & Tool Compatibility

Automatic Reasoning Difficulty Void V1 Flash exposes reasoning_effort natively via the /api/v1/models endpoint. Tools like OpenCode Zen, Claude Code, Cursor, and Windsurf automatically detect and surface a reasoning difficulty selector, no manual configuration needed.

Level

Behavior

low

Fast, minimal reasoning. Best for simple Q&A.

medium

Balanced. Recommended for most tasks.

high

Deep reasoning. Best for complex code & math.

Pass it manually

curl https://0vai.vercel.app/api/v1/chat/completions \
  -H "Authorization: Bearer $VOID_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"messages":[{"role":"user","content":"Think deeply"}],"reasoning_effort":"high","stream":false}'

fetch('https://0vai.vercel.app/api/v1/chat/completions', {
  method: 'POST',
  headers: {
    'Content-Type': 'application/json',
    'Authorization': 'Bearer ' + VOID_API_KEY
  },
  body: JSON.stringify({
    messages: [{ role: 'user', content: 'Think deeply' }],
    reasoning_effort: 'high'
  })
});

import os, requests

r = requests.post(
    "https://0vai.vercel.app/api/v1/chat/completions",
    headers={"Authorization": f"Bearer {os.environ['VOID_API_KEY']}"},
    json={
        "messages": [{"role": "user", "content": "Think deeply"}],
        "reasoning_effort": "high",
        "stream": False
    }
)
print(r.json())

Compatible tools read reasoning_effort_levels from /api/v1/models and surface a difficulty picker automatically. No extra config needed.