Primary use stated or inferred cautiously from official documentation.
SingProbe Ling-3.1-flash-1019
A streaming safety probe reading the base model’s hidden states to score eight query intents, unsafe-response risk and hallucination per token. Its BF16 weights were published on November 19; it is not another generative assistant.
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 | inclusionAI/Ling-3.1-flash-1019-singprobe |
|---|---|
| Model type | Text classification |
| Access model | Open weights |
| License | MIT |
| Deployment | Self-hosted |
| Release | 19 NOV 2026 |
| Knowledge cutoff | Not published |
| Entrance | Base-model hidden states |
| Exit | Risk scores |
| Context window | Probe window: 2,048 tokens |
| maximum output | Ten scores per token |
| 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 · SGLang |
Safety monitoring during generation by the compatible base checkpoint.
Requires the exact base checkpoint and adapted SGLang; the probe window is 2,048 tokens, not the base model’s context. MIT covers the probe; the base retains its terms. Calibrated 0.5 thresholds do not guarantee error-free operation.
What it can do
Orientation
Safety monitoring during generation by the compatible base checkpoint.
Context
Probe window: 2,048 tokens · Ten scores per token
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 |
No verified official API tariff. Self-hosting requires memory, compute and maintenance; downloadable weights do not make inference free.
i Prices change and may depend on level, region or context length. Check the source before making a decision.
How to read the results
Query F1 over six benchmarks and response-safety F1 over eight, equally weighted by benchmark with one threshold selected per suite. Hallucination AUC covers six benchmarks. These are not universal probabilities or Inferama evaluations.
| Benchmark | Result | Metric | Source |
|---|---|---|---|
| SingProbe query intent 6-benchmark suite | 0.8687 | F1 | View source ↗ |
| SingProbe response safety 8-benchmark suite | 0.8707 | F1 | View source ↗ |
| SingProbe hallucination 6-benchmark suite | 0.8041 | AUC | View source ↗ |
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
inclusionAI/Ling-3.1-flash-1019-singprobe
BF16 checkpoint.