Isorun

🚨 Benchmark Friday 🚨 Take a look at these Sandbox Benchmarks from @isorun, @northflank, and newly added @LightningAI. more results in the 🧵

Read more about our persistent storage: docs.isorun.ai/sandboxes/pers…

The architecture is deliberately simple: disposable compute, persistent state when you need it. Excited to build this with @ovaistariq and the Tigris team!

Isorun launched just over a week ago, and today we're partnering with @TigrisData to bring persistence to Isorun agentic runtime. Isorun gives AI agents fast, hardware-isolated runtime for running AI generated code in isolation; Tigris gives them durable, forkable disks.

@pucchkaa 👋 Our 1.26s on the first slice of requests was the worst-case tail, as our capacity-aware load balancer was still settling under the synchronized burst. It self-rebalances the load as the run progresses, which is exactly why the latency drops across the quartiles.

Went through the results and isorun almost crushed the competitors. Northflank and isorun seems like best in class compared to others in terms of total time, p50/p99 for readiness and allocate. Surprised modal had higher values and near consistent across all three percentiles. While northflank and particularly isorun had near subms values in final quarter. What sorcery is this? How do you drop from 1.26s (first quarter) -> 152 ms(last quarter) in latency

🚨🚨🚨 Huge congrats to @isorun for completing the 100k Scale Invitational. See the full results below:

@computesdk @namespacelabs 🚀

🚨 Benchmark Friday 🚨 After weeks of work from us and the participants, today we are excited to release the results of the 2026 100k Scale Invitational. Over the next hour we'll be releasing each set of results. Huge shoutout to @namespacelabs for powering our benchmarks!