Streaming Code World Model generates causally, but its repository explicitly says the reported implementation is not yet real time. Compare generation time with playback duration before repeating the word live.
What Streaming Code World Model actually reports
The repository describes a four-step model producing a roughly 0.71-second chunk in about 5.1 seconds on one H200. It reports a 141 GB H200 configuration and says 80 GB GPUs are not yet supported. Its released adapters depend on an underlying Code World Model/MiniMax-H3 stack.
What this changes for the reader
Using the repository's figures, 5.1 divided by 0.71 is about 7.2: that generation interval is several times longer than playback. This is editorial arithmetic from reported inputs, not a new benchmark. Causality means future frames are not required; it does not imply the machine keeps pace with a player.
A practical evaluation before adoption
Keep chunk playback duration and wall-clock generation duration in separate columns. Compare the initial chunk with later chunks as history grows. Record precision, peak device memory and the complete model dependency set. If evaluating a modified implementation, label it separately instead of treating the original timing as inherited performance.
- Calculate generation-to-playback ratio.
- Record exact H200 configuration and precision.
- Verify base assets as well as adapter files.
Limits and availability to keep in view
We have not reproduced the reported timing or peak memory. The repository's sampled clips and limited evaluation are not proof of arbitrary scene reliability. Public source code also requires a separate review of every model dependency's terms.
Where to go next
Read our WorldPlay2 controls and deployment checks for the adjacent workflow decision. Keep this project's conditions separate from other models' results. If a dependency, artifact or service limit is unclear, leave it unresolved in the comparison rather than filling the gap with an assumption.
Frequently asked questions
Is this project already real time?
The repository explicitly describes it as not yet real time.
Can its published recipe run on an 80 GB GPU?
The repository says 80 GB GPUs are not yet supported.
Has FyreLinkz independently tested Streaming Code World Model?
No. This article reports the linked primary source and proposes an editorial evaluation; it contains no independent performance benchmark.
Sources & further reading
Primary sources checked Oct 18, 2026. Vendor statements are attributed; editorial advice is our own.
- 1Streaming Code World Model — primary source and release details ↗Streaming Code World Model project
Help us keep this useful. Check contact availability →
