GigaAlgo // Neural Systems

Intelligence, engineered at frontier scale.

One API for every model we build — language, vision, audio, code and autonomous agents. Ship frontier capability in an afternoon, not a quarter.

Try the playground Explore capabilities

SOC 2 Type II 99.99% uptime SLA 38 regions

giga-cli · v3.4

Edge · iad1 giga-3-opus

Foundation models

Tokens / second

p99 latency

Uptime SLA

Capabilities

Every modality, one substrate.

Six production surfaces built on a shared training stack. Mix them in a single request, or call one in isolation — the interface is identical.

Language reasoning

Long-context synthesis, structured extraction and multi-step reasoning across a 2M-token window.

2M ctxJSON modetools

Vision & documents

Detection, OCR, layout understanding and video frame analysis returned as typed bounding regions.

detectionOCRvideo

Speech & audio

Streaming transcription with diarisation, plus low-latency neural voice synthesis in 32 languages.

diarisestreamTTS

Code synthesis

Repository-aware generation, review and migration across 40+ languages, with test scaffolding included.

repo-aware40+ langs

Vector embeddings

Multilingual retrieval embeddings with 18 ms median search across billion-scale indexes.

18msmultilingual

Autonomous agents

Planning, tool use and self-correction loops with sandboxed execution and full step tracing.

planningsandboxtrace

Playground

Run it before you read about it.

Pick a capability, edit the prompt, and watch the model respond. This demo streams a representative trace — wire up the real endpoint with the key below.

Prompt

giga-3-opus

Idle · awaiting prompt

REST + gRPC SDK: Python · TS · Go · Rust Streaming tokens Bring your own keys VPC peering

Model catalogue

Five families, one endpoint.

Route by capability, latency budget or cost. Every model shares the same request schema, so switching is a one-line change.

Pricing per 1M tokens · measured at 512-token completions
Model	Modality	Context	Price / 1M	p99 latency
giga-3-opus flagship	Text · Tools	2.0M	$18.00	412 ms
giga-3-sonnet	Text · Tools	1.0M	$6.00	208 ms
giga-3-haiku	Text	256K	$0.90	74 ms
giga-vision-x	Image · Video	512K	$11.00	260 ms
giga-embed-4	Embeddings	32K	$0.12	18 ms

How it works

From raw corpus to production endpoint.

Four stages, fully instrumented. Bring your own data or start from a base checkpoint — the pipeline is the same either way.

Ingest

Stream documents, media and telemetry into governed stores with automatic deduplication and lineage.

Train

Distributed pre-training and LoRA fine-tuning on dedicated clusters, checkpointed every 500 steps.

Evaluate

Regression suites, red-teaming and drift monitors gate every promotion — no model ships unmeasured.

Deploy

Ship to managed endpoints in 38 regions, or export weights to your own VPC and hardware.

In production

Where GigaAlgo already runs.

Four sectors, four very different failure modes. The models hold up because they were evaluated against each one.

A vehicle driving on a city street, used as a mobility deployment example — Autonomy Perception at 60 fps Sensor fusion and edge inference for fleets running without a network round-trip.

Medical radiation therapy machine with a patient table in a clinic, used as a healthcare deployment example — Medical imaging Reads that assist, not decide Segmentation and anomaly pre-screening with clinician-in-the-loop audit trails.

A factory floor filled with industrial machines, used as a manufacturing deployment example — Robotics Visual QA on the line Defect detection trained per production run, retrained nightly on the day's rejects.

Performance analytics graphs displayed on a laptop screen, used as a financial services deployment example — Markets Signals from filings Structured extraction across disclosure archives, with every claim traced to a source line.

Pricing

Metered, not gated.

Start free, pay per token, and stop paying the moment you stop shipping. No seat minimums, no platform fee.

Starter

$0/ month

For prototypes and evaluation. Full model access, rate-limited.

1M free tokens monthly
All five model families
Community support

Create account

Scale

$499/ month

For teams in production. Predictable throughput and real support.

Volume token pricing
Fine-tuning & private endpoints
99.9% uptime SLA
Priority queue & support

Start with Scale

Enterprise

Custom

For regulated deployments and dedicated capacity.

Reserved GPU clusters
VPC peering & data residency
SOC 2 · HIPAA · GDPR pack
Named solutions engineer

Talk to us

Questions

Answers, before the call.

Still stuck? The contact form goes straight to an engineer, not a queue.

Yes. Upload a corpus or point us at your object store, and LoRA fine-tuning runs on dedicated clusters with checkpoints every 500 steps. Your data is never used to train shared models, and you can delete it on demand.

Every model family shares one inference substrate, so routing, batching and KV-cache reuse are handled once. Requests land in the nearest of 38 regions, and small models stay warm at the edge for sub-100 ms first tokens.

Prompts and completions are encrypted in transit and at rest, retained only for the window you configure, and excluded from training by default. Enterprise plans add VPC peering and per-region data residency so traffic never leaves your jurisdiction.

Yes. Enterprise customers can export weights and run them on their own accelerators, with the same SDK surface. Air-gapped installs ship as a signed bundle with offline licence validation.

Purely metered. You pay per token with no seat minimums or platform fee, and the free tier never expires. Spend caps and per-key budgets are configurable from the console, so a runaway job can't surprise you.

Get started

Tell us what you're building.

An engineer replies within one business day with an access key and a worked example for your use case.

hello@gigaalgo.ai email Developer Discord 12k members API documentation v3.4