Primary use stated or inferred cautiously from official documentation.
LingBot-Touch-Small
Robbyant JEPA encoder for magnetic, piezoresistive and visuotactile signals. It turns normalized measurements into representations that other models can use to interpret contact during manipulation.
The essential
Maximum input capacity when published by the source.
Documented maximum generation limit.
Verified availability channels.
Primary functional family and verified specialties.
Declared terms for API access, model weights, or self-hosting.
Limits and integration
| API ID | robbyant/lingbot-touch-small |
|---|---|
| Model type | Robotics |
| Access model | Open weights |
| License | Apache 2.0 |
| Deployment | Self-hosted |
| Release | 09 NOV 2026 |
| Knowledge cutoff | Not published |
| Entrance | Normalized tactile signals |
| Exit | Tactile representations |
| Context window | Five-finger sequences |
| maximum output | Vectors with spatiotemporal addresses |
| Reasoning | Not published |
| Published tools | |
| Structured Outings | Not published |
| Batch processing | Not published |
| Prompt cache | Not published |
| Fine-tuning | Not published |
| Verified platforms | Hugging Face · ModelScope · PyTorch |
Extract tactile features using the 21.76-million-parameter HaT-Small encoder. The bundle includes frozen weights and their inference configuration.
Visuotactile input uses scalar measurements derived from the sensor. The bundle accepts no raw RGB images and contains no robot control policy. Integration requires adapting each sensor’s geometry and ranges.
What it can do
Orientation
Extract tactile features using the 21.76-million-parameter HaT-Small encoder. The bundle includes frozen weights and their inference configuration.
Context
Five-finger sequences · Vectors with spatiotemporal addresses
Tools and integration
Not published
Access
Hugging Face · downloadable checkpoint
Documented cost
| Concept | Worth | Unit/condition |
|---|---|---|
| Standard input | No published official API rate | Standard API rate |
| Cached input | Not published | Reading reused prefixes |
| Cache write or storage | Not published | The condition varies by provider |
| Standard output | Self-hosted infrastructure cost | May include reasoning tokens |
| Batch input | Not published | Asynchronous processing |
| Batch output | Not published | Asynchronous processing |
Weights are distributed under Apache 2.0. The developer publishes inference code separately and does not announce a commercial API for these checkpoints.
i Prices change and may depend on level, region or context length. Check the source before making a decision.
How to read the results
The project reports experiments across 26 physical and 12 simulated manipulation tasks. Results for policies that use tactile features do not represent the success rate of the encoder alone.
i Integrated policy results in the technical report.
Inferama only highlights a “best result” when the metric, test set, configuration, and date allow for an equivalent comparison. The supplier's figures are presented as claims from its own source.
Chronology
LingBot-Touch-Small
Encoder weights and inference configuration.