Frontier AI inference. Sovereign. 80% cheaper.

Open-weight models on Australian infrastructure. OpenAI-compatible API. Your data never leaves the country. One line of code to switch.

View PricingRead the Docs

One line to switch

If you are using the OpenAI SDK, migration is a single line change. Your prompts, tools, and response handling stay identical.

migrate.py
# Before: Anthropic / OpenAI
client = OpenAI(base_url="https://api.openai.com/v1")
# After: Continuum (one line changed)
client = OpenAI(
base_url="https://api.continuum.au/v1",
api_key="your_key"
)
response = client.chat.completions.create(
model="deepseek-v4-flash",
messages=messages
)

Why Continuum Inference

Sovereign

Australian infrastructure. No data leaves the country. ISO 27001 and SOC 2 compliant hosting. Meets APRA CPS 234, SOCI Act, and PSPF requirements.

80% cheaper

Frontier-class models at AUD $3–4 per million tokens. Anthropic charges $15–25. OpenAI charges $10–30. Same capability, fraction of the cost.

Drop-in compatible

OpenAI-compatible API. Change your base_url, keep your code. Function calling, JSON mode, streaming, thinking modes. Zero re-architecture.

Frontier models, benchmarked

DeepSeek V4-Flash delivers 85–95% of frontier closed-model capability at a fraction of the cost. Open-weight. MIT licensed. 1M token context.

DeepSeek V4-Flash

AvailableMIT Licence

284B total parameters, 13B active per token. MoE architecture. 1M native context.

1M tokens
Context Window
13B
Active Params
160 GB
Checkpoint
3 levels
Reasoning Modes

Benchmark comparison

BenchmarkV4-FlashClaude Opus 4.6GPT-5.4
SWE-bench Verified79.0%80.8%~78%
LiveCodeBench91.6%88.8%—
MMLU-ProStrongStrongStrong
Context Window1M~200K1M
Full benchmark details →

The cost comparison

Same API format. Same capability tier. Dramatically different price.

Anthropic

$15–25

AUD per 1M tokens

  • Frontier capability
  • US infrastructure
  • Proprietary
Recommended

Continuum

$3–4

AUD per 1M tokens

  • Frontier-adjacent
  • Australian sovereign
  • MIT licensed

OpenAI

$10–30

AUD per 1M tokens

  • Frontier capability
  • US infrastructure
  • Proprietary

A team spending $20,000/month on Anthropic Claude

saves $192,000 per year

with equivalent capability on sovereign Australian infrastructure

Three steps. Five minutes.

01

Get an API key

Sign up and receive your API key immediately. No credit card required for the first 1M tokens.

02

Change your base_url

Point your existing OpenAI SDK client at api.continuum.au. One line of code. Everything else stays the same.

03

Start saving

Your existing application works immediately. Same request format, same response format, 80% lower cost.

Full API feature parity

Function calling

Up to 128 tools per request. Parallel calls. Strict schema validation. OpenAI-compatible format.

Streaming

Server-sent events streaming. Identical to OpenAI streaming format. Real-time token delivery.

JSON structured output

Native JSON mode via response_format. Guaranteed valid JSON for structured extraction pipelines.

Thinking modes

Three reasoning levels: Non-think (fast), Think High (analytical), Think Max (deep reasoning). Per-request.

1M token context

Native million-token context window. No sliding window. No special configuration required.

Stateless inference

No prompts or completions stored. No training on your data. Process and forget. Zero data retention.

Built on proven infrastructure

H200
NVIDIA GPUs
141 GB HBM3e
1M
Token context
Native, no tricks
99.5%
Uptime SLA
Standard tier
0
Data retained
Stateless inference

Ready to cut your inference bill?

Same models. Same API. Sovereign infrastructure. 80% cheaper. Start in five minutes.

View PricingRead the Docs