Model speed now comes from the serving stack, not the model. vllm's september 13 update for kimi k3…
model speed now comes from the serving stack, not the model. vllm's september 13 update for kimi k3 cut latency 56-60%, lifted throughput 2.2-2.8x and slashed time-to-first-token 72-85%.
the inference engine just became the fastest way to get faster.
Context
vLLM's blog of 13 September 2026 on Kimi K3 performance reports 56 to 60 percent lower latency, 2.2 to 2.8x throughput and 72 to 85 percent lower time to first token across concurrency 1, 4 and 16, on an 8K input and 1K output workload, TP8, eight-token DSpark speculation and a B300 node. The baseline is vLLM v0.27.1 against main commit 82a85dc1, for the same model.
The figures match the first-party post and are vendor-reported, not independent. This is a software-version comparison for one model on one hardware tier, and the update is a main-branch commit and not a tagged release as far as the text says. Other workloads saw smaller per-PR gains of 5 to 41 percent. The inference engine just became the fastest way to get faster is the author's take.
Watch next
- The release tag containing these changes and independent reruns.
Sources
Provenance
The note above is reproduced unedited from the original post, first published on Threads on 20 September 2026 at 21:51 IST. Sources are the papers and datasets the note draws on.
View the original post ↗Embed this note
More notes
The air is now being asked to keep its own ledger
the air is now being asked to keep its own ledger: ecmwf’s aifs compo becomes the first ai model to forecast atmospheric composition globally every three hours, cleanair simulates 365 days of pm2.5 over china in ten seconds, and a unified framework maps six pollutants at one kilometer across the whole country. the air now files its own composition report.
read the note →The current is now being asked to draw its own map
the current is now being asked to draw its own map: china’s langya 2.0 predicts six ocean phenomena including internal waves and mesoscale eddies, a deep net called wenhai resolves eddies globally with air sea flux formulas built in, and scripps infers surface currents from the way temperature patterns deform in satellite images. the ocean now files its own circulation report.
read the note →The soil is now being asked to report its own carbon
the soil is now being asked to report its own carbon: a nix color sensor paired with generative data augmentation predicts soil organic carbon without a lab, random forest drives 74 percent of soil health mapping studies, and sentinel 2 tracks five year carbon change across france and italy from 922 samples. the dirt now files its own carbon account.
read the note →