INDEPENDENT LLM PROVIDER · THAILAND

OpenThai 2.0.1 · Qwen3.8-27B — a self-hosted lab LLM, free for research.

OpenAI- and Anthropic-compatible endpoints on my own GPUs. 262,144-token context, reasoning on, vision in. Free for research and experimentation — invite-only. It runs on a home-lab box, so expect downtime.

free · research useinvite-only262,144 ctxINT8 W8A8thinking · medium effortvisionOpenAI + Anthropic APISSE streaming

Try it free Request access

curl · https://aipassport.trirat.co/v1

curl https://aipassport.trirat.co/v1/chat/completions \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"model":"qwen3.8-27b","stream":true,
       "messages":[{"role":"user","content":"สวัสดี — what can you do?"}]}'

Base URL https://aipassport.trirat.co/v1 · model qwen3.8-27b · streaming recommended. Anthropic SDKs work too — same base URL, call /messages with x-api-key. Wiring an agent? The API reference has measured limits and ready-made configs for Claude Code, opencode, pi, Hermes and OpenClaw.

Live performance

2.50s

Latency · p50 TTFT

6055 tok/s

Prefill · tok/s · 24h

57.1 tok/s

Throughput · decode

93.0%

Uptime · 30d

Measured from real traffic (Prometheus), 24h window · uptime 30d.

Free for research and experimentation

No pricing, no cards, no baht. This is a personal lab LLM — free to use for research, experiments, and tinkering. Access is invite-only.

Free · invite-only

฿0 — always

✓Full model access — 262,144-token context, reasoning, vision
✓OpenAI- and Anthropic-compatible endpoints
✓For experiments and research — not for production

Request access

Heads up — it's a lab box

✓Runs on my own GPUs at home
✓Usually up ~07:00-18:00 ICT (Bangkok); trains overnight ~18:00-07:00
✓Best-effort, no SLA - the schedule can change without notice

If it's unreachable it's probably mid-training - try again during 07:00-18:00 ICT (Bangkok), or DM me anytime and I can spin it up for you.

Model

OpenThai 2.0.1 · Qwen3.8-27B (id qwen3.8-27b)

Precision

INT8 W8A8

Context

262,144 tokens

Reasoning

on — medium effort

Vision

yes

API

OpenAI /v1/chat/completions · Anthropic /v1/messages

Streaming

SSE

Max output

32,768 tokens

Reference

Full API spec + harness setup →

Request access

Access is invite-only and free for research and experimentation. Message me — tell me what you want to build or test. Keys are issued by hand, usually within a day.

Message on Facebook Email [email protected]

Free · invite-only · best-effort (expect downtime).