Hi Liberty Token Factory · Built for Southeast Asia

Run any model.
Answer to no one.

Open-source inference at production scale, served entirely from within Southeast Asia — dedicated endpoints, transparent per-token pricing, and zero vendor lock-in for teams building in Singapore, Indonesia, Thailand, Vietnam, Malaysia and the Philippines.

70+
open models, one API
99.95%
uptime SLA
120M+
tokens per minute
<400ms
p50 latency, verified
Inference

Open models, closed gaps.

Deploy Llama, Qwen, DeepSeek and GPT-OSS on dedicated endpoints with sub-second targets and 99.95% uptime. Autoscaling, speculative decoding and routing across our Southeast Asia data centers hold latency steady from your first request to your hundred-billionth.

Explore endpoints →
deepseek-v3.2 · dedicatedsingapore
llama-4-scout · sharedjakarta
qwen3-235b-a22b · dedicatedbangkok
gpt-oss-120b · sharedautoscaled · SEA
bge-multilingual-gemma2embeddings
Why teams switch

Everything a production inference stack needs, nothing you have to build yourself.

Scale without a ceiling

Move from a single prototype to hundreds of millions of tokens a minute without re-architecting. Autoscaling and routing across our Southeast Asia endpoints hold latency steady the whole way.

Pricing you can predict

Transparent per-token rates across shared and dedicated tiers, with volume discounts that apply automatically as usage grows. No idle GPU charges.

State-of-the-art multimodal models

Text, code, vision and reasoning models behind one API. Mix modalities in the same pipeline without juggling providers or contracts.

Agent-ready primitives

Native function calling, structured JSON output and built-in safety guardrails, so agents behave predictably once they're talking to the real world.

Bring your own fine-tune

Adapt a base model to your data, then deploy the checkpoint straight to a Hi Liberty endpoint with the same per-token pricing and SLAs as everything else.

RAG, batteries included

High-performance embedding models and vector storage sit next to inference, so indexing, retrieval and generation stay inside one governed platform.

The platform, broken down

Three services, one account.

Inference Service

Serve, at production SLAs

Text, code and vision models behind a single OpenAI-compatible API, backed by production SLAs and predictable per-token pricing.

Learn more →
Data Lab

Turn logs into training data

Explore inference history, curate datasets from production traffic, and move straight into your next fine-tuning pass.

Learn more →
Post-training

From dataset to deployed

Streamlined fine-tuning workflows that take an open model from your data to a live, billable endpoint in one pass.

Learn more →
Model catalog

Top open-source models, ready today.

Text & multimodal

DeepSeek-V3.2 Llama 4 Scout Qwen3-235B-A22B GPT-OSS 120B & 20B Mistral Large 3 Gemma 3 27B Kimi K2 GLM-4.6

Embeddings & guardrails

BAAI/bge-en-icl BAAI/bge-multilingual-gemma2 intfloat/e5-mistral-7b-instruct Qwen3-Embedding-8B Llama-Guard-4-8B
View all models →
Benchmark-backed

Performance and cost efficiency, verified by someone other than us.

Verified, not claimed

Independently benchmarked throughput and latency across flagship models, so procurement doesn't have to take our word for it.

Built to hold the line

A 99.95% uptime SLA, autoscaling and speculative decoding keep response times flat under real production load, not just in demos.

70+ models, one bill

LLMs, vision, reasoning and embedding models under a single account and a single invoice — expanding every month.

Familiar API

If you know the OpenAI SDK, you already know this.

Point your existing client at Hi Liberty and switch models with a string, not a migration.

Read the docs →
from openai import OpenAI

client = OpenAI(
    base_url="https://api.hiliberty.ai/v1",
    api_key=HI_LIBERTY_API_KEY
)

completion = client.chat.completions.create(
    model="deepseek-v3.2",
    messages=[{
        "role": "user",
        "content": "What does liberty mean for infrastructure?"
    }]
)

print(completion.choices[0].message.content)
# -> "Owning the stack you build on."

Hi Liberty pricing

Start on shared access, move to dedicated endpoints with reserved capacity and a 99.95% SLA whenever you need it. Transparent $/token throughout, no idle infrastructure costs.

Questions

Questions and answers about Hi Liberty Token Factory

Can I use this for large production workloads?

Yes. Hi Liberty is built for large-scale, production-grade AI workloads. Dedicated endpoints deliver sub-second inference, a 99.95% uptime SLA, and autoscaling throughput for workloads well past a hundred million tokens a minute.

Scale from experimentation to full production across our Southeast Asia footprint, without rate throttles or GPU management of your own.

I'd like to use a model you don't support yet — what do I do?

We onboard new open-source releases regularly, based on customer demand and independent benchmarking. Enterprise users can request model optimization or custom deployment support directly through our Solutions team.

Can I get a dedicated instance?

Yes. Dedicated endpoints provide guaranteed isolation, predictable latency and reserved compute capacity, including a 99.95% SLA, custom autoscaling and a choice of deployment location within our Singapore, Jakarta or Bangkok facilities. Contact our team to size an instance for your workload and compliance needs.

How do I deploy my own fine-tuned model?

Deploy directly through the dashboard or API. Each deployment runs with transparent per-token pricing and inherits the same security and latency guarantees as standard endpoints. Full post-training workflows let you go from dataset to deployed checkpoint in one pass.

How secure is the service, and where does my data go?

A zero-retention mode is available, where requests and outputs are never stored or reused for training. Data is processed exclusively within our Southeast Asia facilities (Singapore, Jakarta and Bangkok), certified to SOC 2 Type II and ISO 27001, and never leaves the region.

What are the rate limits, and can they be increased?

The free tier ships with generous defaults; enterprise plans remove caps entirely. We size endpoints to your traffic profile — reach out and we'll scope an endpoint for your real-world workload.

Which countries can I use Hi Liberty from?

Hi Liberty currently serves customers based in Southeast Asia only — including Singapore, Indonesia, Malaysia, Thailand, Vietnam and the Philippines. All infrastructure, billing and support are local to the region. We don't yet support sign-ups or billing outside Southeast Asia; reach out if you'd like to be notified when that changes.

How does pricing compare to other providers?

Pricing is transparent per-token, with clear input/output separation and volume discounts as you scale. No hidden infrastructure or idle GPU costs — you pay only for what you serve.

Start your journey

Build on infrastructure
that answers to you.