A
APPLE / apple/FastFrames-L

FastFrames L

A 14B-parameter video model derived from Wan2.1-14B, with a public checkpoint dated November 20. It combines a two-step causal generator, pixel decoder and KV-cache refinement to reduce latency. The recipe generates 480p video from text.

01 / SUMMARY

The essential

ORIENTATIONTwo-step autoregressive video diffusion

Primary use stated or inferred cautiously from official documentation.

CONTEXTNot published

Maximum input capacity when published by the source.

EXIT832 × 480 · 81 frames · 16 fps

Documented maximum generation limit.

ACCESSHugging Face · downloadable checkpoint

Verified availability channels.

MODEL TYPEVideo

Primary functional family and verified specialties.

LICENSEApple AMLR · non-commercial research

Declared terms for API access, model weights, or self-hosting.

02 / TECHNICAL SHEET

Limits and integration

API IDapple/FastFrames-L
Model typeVideo
Access modelOpen weights
LicenseApple AMLR · non-commercial research
DeploymentSelf-hosted
Release20 NOV 2026
Knowledge cutoffNot published
EntranceText
ExitVideo
Context windowNot published
maximum output832 × 480 · 81 frames · 16 fps
ReasoningNot published
Published tools
Structured OutingsNot published
Batch processingNot published
Prompt cacheNot published
Fine-tuningNot published
Verified platformsHugging Face
BEST SUITED FOR

Research into low-latency video generation.

WORTH MONITORING

Weights use Apple AMLR for non-commercial research; code and dependencies have separate licenses. FastFrames S’s 36 ms result and the LingBot demo are not FastFrames L measurements. The L profiling recipe uses eight GPUs.

03 / CAPABILITIES

What it can do

01

Orientation

Research into low-latency video generation.

02

Context

Not published · 832 × 480 · 81 frames · 16 fps

03

Tools and integration

Not published

04

Access

Hugging Face · downloadable checkpoint

Input modalities
Text
04 / PRICES

Documented cost

ConceptWorthUnit/condition
Standard inputNo published official API rateStandard API rate
Cached inputNot publishedReading reused prefixes
Cache write or storageNot publishedThe condition varies by provider
Standard outputSelf-hosted infrastructure costMay include reasoning tokens
Batch inputNot publishedAsynchronous processing
Batch outputNot publishedAsynchronous 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.

05 / EVALUATIONS

How to read the results

“

No verified quantitative score for this version.

i No verified quantitative score for this version.

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.

06 / VERSIONS

Chronology

apple/FastFrames-L

Version added to Inferama's verified catalog.

07 / SOURCES

Traceability