Models · Yours, not rented

Models that give your people their time back.

Trained on your own data, they take over the work that piles up: A&D diligence, month-end close, ticket coding, regulatory filings. Lower admin cost, faster cycles, and your staff back on what matters. You own the weights.

Named for the crews whose busywork they take on.
Sparky7B
REGULATORY · AER DIRECTIVES · LAND FILES
Torquay7B
DRILLING · TOUR SHEETS · DDRS
Gauger13B
PRODUCTION · SCADA · DECLINES · FAILURES
San Andres3BFOUNDATION
OIL & GAS CORPUS · EVERY MODEL STARTS HERE
ONE WELLBORE OF DATA · FOUR MODELS
The catalog

Two ways to get a model.

You don't have to know anything about AI to start. Either we build one that learns your operation, or you pick a proven public model and go today. Both run on your own machines.

01We build it for you

It learns your operation

We take our base model and train it on your wells, tickets, and history until it speaks your operation. It runs on your machines and the weights are yours to keep.

Trained on
Your own field data
Ready in
A few weeks
Best for
Work that must know your basin
Ownership
You keep the weights
Available nowNamed for the crews
San Andres
FOUNDATION3B
Torquay
DRILLING7B
Gauger
PRODUCTION13B
Sparky
REGULATORY7B
COMING SOONRoughneck · rig opsMudlogger · geologyPumper · field
02Start today

Grab a proven model now

Open models we have already checked over: license cleared, weights verified, tested on field work. Self host it on your own hardware, or run the same model through our API. Same weights either way.

INCLUDED
API access comes with your Signal or Fleet Builder seat. The rates opposite apply only if you build outside the harness, on prepaid credits.
ModelSizeContextIn / 1MOut / 1M
DeepSeek R1671B MoE128K$0.55$2.19
Qwen3235B MoE256K$0.22$0.88
Llama 4 Scout109B MoE10M$0.18$0.59
gpt-oss120B128K$0.15$0.60
Mistral Large 3123B256K$0.50$1.50
Gemma 327B128K$0.11$0.19
Devstral 224B256K$0.10$0.30
Qwen38B128K$0.035$0.14
USD per 1M tokens · cost = input × in + output × outEvery model here is downloadable to self-host at no per-token cost
LICENSE REVIEWEDWEIGHTS VERIFIEDBENCHMARKED ON FIELD WORKSIZED TO YOUR HARDWARE
Included

The models come with the harness.

Inference is part of your Signal or Fleet Builder subscription, not a separate line item. Pick the model that fits the job and put it to work — you buy a plan size up front, not a meter that runs. Individual is the baseline; Professional and Team come at 5× or 20×, and every extra seat adds another baseline to the shared pool.

BILLINGper plan, per month
MODELSthe whole library, switch freely
INFERENCE1× · 5× · 20× plan sizes
EGRESSno charge
SWITCHINGeither way, prorated
Enterprise

Want the tools running in your environment?

If your agents need to reach into your historian, your land system, or anything else behind your firewall, that is an Enterprise conversation. Single tenant, your own hardware, your own endpoints, priced to the deployment rather than the seat.

Inference and tool execution inside your boundary
Dedicated capacity, no shared pool
Private connectors to systems that never touch the internet
Annual agreement, volume pricing
What it changes

Fewer hours lost to the busywork.

You don't need to know a thing about AI, or about the Montney, to feel the difference. The models take on the repetitive work that eats your team's week, so the people you hired can do the work only they can.

Lower admin cost

Filings, reconciliations, and reports that quietly pile up back-office hours get done automatically.

Maximize staff time

Engineers and operators stop copy-pasting between systems and get back to work only they can do.

Automate the process

From field ticket to posted invoice, or month-end to submission, the routine path runs itself behind a human gate.

Smoother A&D

Data-room diligence, well-file abstraction, and reserves prep in days instead of weeks.

ADAPTED TO YOUR BASINMontneyDuvernayCardiumClearwaterViking+ yours
The routing harness

The hardest questions get the smartest model. Your edge stays safe from competitors.

You don't need to know anything about AI. We set up the routing for you: everyday work runs on models inside your walls, and when a question is too hard for them, it goes to the most capable model in the world with your business data stripped out first, so nothing that makes you money ever reaches a rival's training set.

APTURE ROUTING
ALPHA RETAINED
The harness
Wrapped before it runs

Before any task runs, the harness checks the model is capable of it, then puts guardrails, approvals, and an audit trail around the work.

In reach
95% stays on your metal

Everyday work runs on your open-weight model with the real data attached. Nothing leaves the building.

Too hard
The 5%, scrubbed first

Only the toughest questions escalate, and only after the scrubber strips every name, coordinate, and volume. The frontier sees a riddle, never your alpha.

THE HARNESS PROVES THE MODEL CAN DO THE JOB. THE SCRUBBER MAKES SURE YOUR ALPHA NEVER LEAVES.

Not sure where to start?

Tell us where the hours go. We find the busywork worth removing and build a model for it, on your data, on your machines.

Start a custom model