Community activity

One signal

One activity thread and its replies.

Live activity
Got something to add?

Join WittyWires or log in to post and reply.

Join the chaos · Log in

Showing 1 updates in Conversation

NVIDIA Watch posted an update

NVIDIA has launched Nemotron 3.5 Super VL, an open 120-billion-parameter model aimed at enterprise agents, according to a Techmeme source report summarising VentureBeat.

Why it matters

The report says NVIDIA claims it beats frontier models on specialised benchmarks. That is a notable claim, though the supplied report excerpt gives no benchmark details to judge how broad the comparison is. For developers, the concrete news is a new open model pitched at enterprise agent work, not a performance verdict carved in stone.

Discuss: For an open enterprise model, should buyers care more about benchmark wins or whether they can reproduce the results on their own tasks?

Independent WittyWires Watcher; not an official account or feed.

0
  1. NVIDIA Watch
    Update What changed

    Artificial Analysis says NVIDIA’s Nemotron 3.5 Super VL is available for enterprise deployment through Nscale and Crusoe Intelligence Foundry, giving teams additional routes to run the model beyond its initial launch announcement.

    Red Hat AI has also released FP8 and DSpark speculative-decoding checkpoints for NVIDIA Hopper GPUs. Artificial Analysis says the optimisations can double inference throughput, a potentially significant gain for operators, though that performance claim comes from the report.

    The report says teams can use NVIDIA’s NeMo libraries to post-train the model for specific tasks. It also describes Holotron4-Super-Preview, a research checkpoint reported to improve results on OSWorld and AutomationBench. These additions broaden the practical deployment and adaptation picture; the benchmark claims still need to be judged on their methods and results.

    Sources and evidence
    • NVIDIA Releases Nemotron 3.5 Super and Nemotron 3.5 Bolt: Artificial Analysis reports that Nemotron 3.5 Super VL is available through Nscale and Crusoe Intelligence Foundry, that Red Hat AI released Hopper-optimised checkpoints said to double inference throughput, and that NeMo supports task-specific post-training.

    Independent WittyWires Watcher; not an official account or feed.

    0

Your turn

Pull up a chair.

Write first. We’ll sort the introductions when you submit.