Noam Hurwitz

Something new. Previously head of engineering @alchemy.

[Noam Hurwitz @ProbablyNoam]

@gajesh @ivanfioravanti @poolsideai 🔥🔥🔥

[Noam Hurwitz @ProbablyNoam]

@mitch_troy @ankrgyl 🔥🔥🔥

[vLLM @vllm_project]

vLLM hit new peak bs=1 decode on Kimi-K3: 464 tok/s 🚀

Under a low-entropy reasoning workload, Kimi-K3 + DSpark on vLLM reaches 464 tok/s on batch size 1 with 4×4 GB300.

This benchmark is fully reproducible with public image: vllm/vllm-openai:kimi-k3 and @inferact's DSpark draft model linked in the thread.

1/3

[Noam Hurwitz @ProbablyNoam]

@ScottHickle @ThroneScience 🔥🔥🔥

Congrats y'all. Few understand the work that goes in under the seat to get here. My gut thanks you every day.

[Noam Hurwitz @ProbablyNoam]

@0xkydo @poolside This is awesome.

[Noam Hurwitz @ProbablyNoam]

@mercor_ai @FireworksAI_HQ If Opus 5 is ranking at the top of your leaderboard for AI productivity, you may want to revisit your benchmarks.

[Noam Hurwitz @ProbablyNoam]

@devstein64 That makes one of us, my friend.

[Noam Hurwitz @ProbablyNoam]

@shengkun_ye @OctenAI @monid_ai @0x_kyle can you confirm?

[Noam Hurwitz @ProbablyNoam]

@HarukaKunori Correct.

[Nathan Lambert @natolambert]

Kimi K3 license. It's inspired by MIT but distinctly non-commercial, where any company making over $20M/yr must get a specific commercial deal (and display Kimi K3 if over 100M users or $20M/mo revenue)

[Jensen Huang @JensenHuang]

For my first post, I’m sharing a letter @NVIDIA signed on why open models matter.

AI will transform every industry, power every company, and be built by every country.

Open models strengthen safety and cybersecurity, accelerate innovation and diffusion, and enable sovereignty.

The world needs both frontier closed models and frontier open models.

https://t.co/AUKzoQ5Ikb

[Noam Hurwitz @ProbablyNoam]

Feels bigger than the DeepSeek moment. Its clear that benchmarks don't tell the whole story, and that Fable/Sol tokens are likely still higher ROI on many tasks, but still - this is crazy.

[Guillermo Rauch @rauchg]

Kimi K3 is the best performing model on https://t.co/aporqgIfIh, ahead of Fable, reaching a comparable success rate in less time.

This is the first time that an open model is ahead of all proprietary ones for this comprehensive web engineering benchmark.

Notes:

▪️ Benchmarks don’t always tell the full story, although this is important signal, adding to mounting evidence that this could be a breakthrough moment for open models

▪️ No model as of yet has reached 100% completion on this set of evals. The top performer peaks at 92% and 96% “with help”

[Unsloth AI @UnslothAI]

@Kimi_Moonshot Congrats guys on the release and thanks for supporting opensource! 🙏🥰

We'll hopefully be able to make 1-bit GGUF quants so people can run them locally (if you can) 😭

[Noam Hurwitz @ProbablyNoam]

@uriklarman @FalconXGlobal Congrats!

[Noam Hurwitz @ProbablyNoam]

@andrewmccalip Jane Street, RenTech, etc already benchmark on this. The frontier labs no one talks about.

[Noam Hurwitz @ProbablyNoam]

@MechPete @heelbitee big facts.

[Noam Hurwitz @ProbablyNoam]

@0xBADB01E OSS models provide an equalizer here, increasingly don’t have to just rent to frontier labs to turn compute to $ + compute ratio of inference:training will keep increasing.

[Noam Hurwitz @ProbablyNoam]

@omarsayha 🔥🔥🔥

[Noam Hurwitz @ProbablyNoam]

@glennonchain https://t.co/Bwo86tVGqp

[Jon Charbonneau 🇺🇸 @jon_charb]

many such cases