←Search results

By relevance · Total 351 results Clear search

Thursday

  1. Story1 reports · 1 facts

    Ben Horowitz recalls the experience of Loudcloud's bandwidth costs plummeting

    Ben Horowitz recalled his experience founding Loudcloud in 1999: at the time bandwidth cost $1,500 per megabit and was hard to buy even with money, forcing the company to sign three-year high-priced long-term contracts; just two years later bandwidth prices fell to $50, and companies that borrowed money to stockpile bandwidth were crushed by the cycle. He uses this to contrast with the current AI application layer, where the price per million Token calls keeps plunging, while prices for GPUs, data center land, power quotas and skilled workers are rising across the board, arguing that the high rental premiums of compute cloud vendors are inherently fragile. He also judges that when on-device small models catch up with large models in performance, the blind worship of cloud-based thousand-GPU and ten-thousand-GPU clusters may cool down within a few months.

    Ben Horowitz recalled his experience founding Loudcloud in 1999: at the time bandwidth cost $1,500 per megabit and was hard to buy even with money, forcing the company to sign three-year high-priced long-term contracts; just two years later bandwidth prices fell to $50, and companies that borrowed money to stockpile bandwidth were crushed by the cycle. He uses this to contrast with the current AI application layer, where the price per million Token calls keeps plunging, while prices for GPUs, da

    Matched fieldsSummary

Sunday

Friday

  1. Story20 reports · 5 facts

    Nvidia releases Nemotron 3.5 Super VL and Bolt models

    On October 15, 2026, Nvidia announced the launch of Nemotron 3.5 Super VL and the edge-oriented 3.5 Bolt models. According to Nvidia's official introduction, Super VL is a 120B-parameter open model aimed at always-on agent tasks, and the company says it has the highest throughput among comparable open models, 5 times faster than the next-best model; it also launched two edge models, Bolt 74B-A9B 24GB and Bolt E4B 4GB, with the former able to run agent tasks locally on RTX/DGX Spark and the latter aimed at voice and conversational agents on GeForce RTX and Jetson. Super VL and Bolt 74B-A9B are already available on Hugging Face, and Bolt E4B is scheduled for release on October 23. On the same day, SGLang announced Day-0 support for Super VL, reaching 243.9 tok/s on a single RTX 6000 PRO with the help of MTP; Prime Intellect and Nvidia used GRPO to post-train it for about 11 hours on 128 GB200 GPUs, raising its score on the public AutomationBench-AA set from 34.0% to 57.7%. That evening OpenRouter listed the model, and the next day Perplexity announced that Pro and Max subscribers and the Agent API can use it.

    On October 15, 2026, Nvidia announced the launch of Nemotron 3.5 Super VL and the edge-oriented 3.5 Bolt models. According to Nvidia's official introduction, Super VL is a 120B-parameter open model aimed at always-on agent tasks, and the company says it has the highest throughput among comparable open models, 5 times faster than the next-best model; it also launched two edge models, Bolt 74B-A9B 24GB and Bolt E4B 4GB, with the former able to run agent tasks locally on RTX/DGX Spark and the latte

    Matched fieldsSummary

Saturday

Thursday

Wednesday

Sunday

Tuesday

Thursday

Tuesday