Sage-A1.5 / LOCAL INTELLIGENCE

Intelligence.
On your terms.

Sage-A1.5 · 35B-A3B. Run inference locally, connect your tools, and take on search and coding tasks. Keep control of your private context.

Keep private material off the model provider’s servers.
35B / 3BTotal / active parameters per token
LOCALYour device. Your context.
Private files
{ }Tools & code
Local model
Privacy protection
ONE MODEL. MULTIPLE DEPLOYMENT PATHS.

Choose your compute.

01 / MORE THAN A RUNNING MODEL

From local inference
to a completed task.

The model reasons; the workspace acts. Connect the service and tools to bring search, reading and code verification into one workflow.

LOCAL WORKSPACE
RECORDED TASK / REPLAY
A SAVED SUCCESSFUL TASK

How many static methods does Java 17’s Objects class declare?

5 searches · 3 page reads · 1 Python check
LOCAL INFERENCE / TOOL EXECUTION
Sage-A1.5Sage-A1.5127.0.0.1
≡read_pageTarget page request
</>code_toolsLocal execution
00:00 / 00:30
Edited trace + local architecture illustration. Not live inference or a hardware test.

02 / A REAL TASK ON YOUR OWN DEVICE

From model startup.
To an answer.

Watch Sage-A1.5 run on a Mac: submit a search through the local web interface and follow the tool calls through to the answer.

Local search · Recorded demo02:36 · Mac · llama.cpp
Mac · llama.cpp · A local model working with search toolsWatch full screen or seek to any moment

03 / LOCAL INFERENCE. YOUR CONTROL.

Inference stays here.
You choose what connects.

Local deployment removes the need to send model requests to a model provider. Your workspace configuration determines whether tools connect and what they send.

01 / PRIVATE DATA

Private material stays local

Use internal documents, source code and research data for inference on your own device, without uploading them to a cloud model service.

02 / YOUR CONTROL

Control access and retention

Set file access, tool permissions and log storage yourself, so the workspace follows your own data management requirements.

03 / OFFLINE READY

Work with local tools offline

Prepare weights and dependencies, then disable external services to work with local files and code tools offline, reducing outbound data transfers.

No cloud model API needed for inference
YOUR DEVICE / PRIVATE WORKSPACE
Sage-A1.5Local inference

Files · Conversation context · Tool results

∅Model providerNo cloud inference requests

LOCAL MODEL. REAL WORK.

Bring agentic capabilities to your own workspace.

Explore the workspace↗
RECORDED TOOL CALLS

Inspect the evidence at every step.

Original tool requests and responses, without system prompts or internal reasoning. The selected record is marked correct. This page does not rerun the model.