https://ja.mangomans.com/v1
mangoismydaddy
Enter administrative token to access telemetry and playground
Manage the autonomous lifecycle of the Molab GPU notebook session.
Separate dedicated ports for API proxy and administrative dashboard.
https://ja.mangomans.com/v1https://jd.mangomans.com125B Total / ~6Bโ14B Active Parameters MoE with Hybrid Gated DeltaNet Linear Attention
| Context Length | TTFT (Time to First Token) | Prefill Throughput | Decode Speed (tok/s) | KV Cache VRAM |
|---|---|---|---|---|
| 20,000 tokens (20k) | ~7.8 seconds | 2,600 tok/s | 110 โ 135 tok/s | ~2.4 GB |
| 50,000 tokens (50k) | ~21.5 seconds | 2,350 tok/s | 95 โ 115 tok/s | ~5.8 GB |
| 100,000 tokens (100k) | ~48.0 seconds | 2,050 tok/s | 80 โ 98 tok/s | ~11.5 GB |
| 256k tokens (Max Window) | ~165 seconds | 1,550 tok/s | 65 โ 80 tok/s | ~29.2 GB (Host Swap) |
Connect any standard OpenAI client (Cursor, OpenWebUI, Python, Node.js) to your dedicated domain.
https://ja.mangomans.com/v1https://jd.mangomans.commangoismydaddydealignai/Qwen3.8-Flash-Next-ABLITERATED-NVFP4qwen3.8-flash-next, defaultcurl -N -X POST https://ja.mangomans.com/v1/chat/completions \
-H "Authorization: Bearer mangoismydaddy" \
-H "Content-Type: application/json" \
-d '{
"model": "dealignai/Qwen3.8-Flash-Next-ABLITERATED-NVFP4",
"messages": [{"role": "user", "content": "Hello!"}],
"stream": true
}'
from openai import OpenAI
client = OpenAI(
base_url="https://ja.mangomans.com/v1",
api_key="mangoismydaddy"
)
response = client.chat.completions.create(
model="dealignai/Qwen3.8-Flash-Next-ABLITERATED-NVFP4",
messages=[{"role": "user", "content": "Write a fast Go concurrent pipeline."}],
stream=True
)
for chunk in response:
content = chunk.choices[0].delta.content or ""
print(content, end="", flush=True)
Connecting to gateway stream...