Bring examples of the task you want done. Train a model, check the improvement on examples kept out of training, and download the result. Review the quote before you start.
New workspaces start with $10 of credit while launch seats remain. No card needed for the first run.
Cell 2 log: axis 3 torque spiked to 118 percent for 400 ms.joint_overtorque
Example rows from robot-faults-eval.jsonl, each resolving to one supported fault label.Free-text logs from factory equipment go in. The tuned model answers with exactly one of four supported fault labels, a fixed string a downstream system can route on instead of prose it has to parse.joint_overtorquevision_dropoutgripper_slipsafety_stop
Explore larger models or start with a measured example. Farka checks which hardware fits your training setup before quoting the run.
80 GB accelerator memory class. Fit depends on the full training configuration and available capacity.
16 GB accelerator memory class. Fit depends on the full training configuration and available capacity.
Compare before and after on the same examples kept out of training. These are internal classification tests, not a guarantee of results on your data.
Correct answers before and after training
Light tune · 1 pass · 16 GB accelerator
270 training examples and 30 test examples. Internal classification task.
Base and trained model tested on the same 30 examples kept out of training.
Read the evidenceCorrect answers before and after training
Full tune · 3 passes · 16 GB accelerator
600 training, 60 development and 60 blind test examples. Internal classification task.
Base and trained model tested on the same 60 examples kept out of training.
Read the evidenceBring examples of the task and the answers you want. Farka checks their format, quality and duplicates before training.
Choose a supported model and review the quote. Train from labeled examples or define outcomes that can be checked automatically.
Compare the model before and after training on the same test examples. See what improved and what still needs work.
Once the run passes its quality checks, download model files or a package ready to run on your infrastructure.
Train from labeled examples, or let clear checks score each answer. Test the result on examples kept out of training before delivery.
Ingest, inspect, and clean your data. Token counts, label balance, and quality signals surface before you spend a credit.
Plan distinct specialists, test a small pilot, and prepare the next group once its results meet your requirements.
Build and probe retrieval indexes over your corpora, and wire them into runs and playground sessions.
Export finished models as self-hosting bundles. Managed endpoints appear only when enabled for your deployment.
Prepaid, usage-based credits. See exactly what each run costs, metered from provider compute evidence, with no surprise invoices.
Pin a training run to the EU or the US and it runs nowhere else. Choose the infrastructure path, retention window, and delivery target. Delete generated datasets immediately, request a portable export, or start a tracked user or workspace deletion process.
Pinning is enforced, not advisory: a run that cannot be placed in-region fails rather than moving. Available on deployments whose control plane, database, and storage are all declared in that region.
Your quote covers training, data storage, quality checks and your chosen delivery option. Review the cost before you start.
Bring a larger model, unusual objective, private infrastructure boundary, proprietary data format, or tailored deployment requirement. We will scope a custom training and evaluation path around the outcome your company needs.
Create a workspace, upload a dataset, and quote your first run before any GPU starts.
Create your workspace