Primary use stated or inferred cautiously from official documentation.
Depth Pro 2 L
Apple family released on November 20 that estimates per-pixel metric depth from photos and video. It combines attention across frames with a decoder per image to preserve edges and temporal consistency. Each size has its own weights; the focal estimator is a separate component.
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 | apple/DepthPro2-L |
|---|---|
| Model type | Vision and geometry |
| Access model | Open weights |
| License | Apple AMLR · non-commercial research |
| Deployment | Self-hosted |
| Release | 20 NOV 2026 |
| Knowledge cutoff | Not published |
| Entrance | Image · Video |
| Exit | Depth maps |
| Context window | Frames at input resolution |
| maximum output | Depth in meters per pixel |
| 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 |
Research into visual geometry, video and scene reconstruction.
Apple AMLR limits the weights to non-commercial scientific research and academic development; it is not Apache 2.0. Infinite values for sky or distant regions are not formatting errors. The focal estimator adds 23.5M parameters to each size.
What it can do
Orientation
Research into visual geometry, video and scene reconstruction.
Context
Frames at input resolution · Depth in meters per pixel
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
Apple performance measurements on one NVIDIA B200, BF16 and uncompiled PyTorch: a 32-frame 1080p clip processed in one call, including the focal estimator. FPS and reserved memory apply to that protocol, not single-frame latency or any GPU.
| Benchmark | Result | Metric | Source |
|---|---|---|---|
| Depth Pro 2 B200 1080p BF16 32-frame throughput | 9.1 | fps · higher is better | View source ↗ |
| Depth Pro 2 B200 1080p BF16 32-frame peak memory | 21.7 | GiB · lower is better | 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
apple/DepthPro2-L
Version added to Inferama's verified catalog.