NvidiaCommercial
Open-weight speech model released on Oct. 28, 2026. Nvidia says it transcribes 40 languages, translates 31 to or from English, and labels up to eight speakers in recordings up to one hour. About 4 billion of its 30 billion parameters are active per token. The available checkpoint is for recorded audio, and Nvidia says quantized local versions are still to come.
Latest update Oct 28, 2026
01
Weights are on Hugging Face and ModelScope. Live streaming and quantized local builds are not in this release.
02
Nvidia on Oct. 28 released open weights for Nemotron 3.5 Transcribe, a model it says transcribes 40 languages, translates 31 to or from English, and labels up to eight speakers in recordings as long as one hour. The company says about 4 billion of its 30 billion parameters are active per token, and that the release is for recorded audio, not live streams. Quantized local versions are described as coming later. The same release includes Granary v2, whose public manifest Nvidia counts at 463,062 hours across 48 languages, separate from the model's full training mix.
03
Nvidia's shipping flagship: rack-scale AI systems built around Blackwell and Blackwell Ultra GPUs - the GB300 NVL72 links 72 GPUs into one giant accelerator, delivering up to 50x better agentic-AI performance than Hopper per SemiAnalysis benchmarks. GB300 entered mass production in mid-2025 and has been the engine of record data-center revenue since.
Nvidia's next-generation seven-chip platform - Vera CPU, Rubin GPU, NVLink 6 switch, ConnectX-9, BlueField-4, Spectrum-6, and the Groq 3 LPU integrated after a December 2025 licensing deal. Rubin NVL72 racks train mixture-of-experts models with a quarter of the GPUs Blackwell needs and deliver up to 10x inference throughput per watt at a tenth the cost per token.
Blackwell-based gaming and creator GPUs, topped by the RTX 5090 ($1,999 MSRP, launched January 2025). FY2026 gaming revenue hit a record $16.0B, but supply to board partners was reportedly cut ~20% in 2026 to prioritize data-center silicon, pushing street prices 15–19% above MSRP.
DRIVE AGX Thor and DRIVE Hyperion power L4-ready vehicles from BYD, Geely, Isuzu, and Nissan plus Hyundai/Kia and Uber robotaxi programs, while Jetson AGX Thor - the Blackwell-powered 'robot brain' released August 2025 with 7.5x the AI compute of Orin - is the standard compute platform for humanoid-robot makers including Figure, Boston Dynamics, Agility, and Amazon Robotics. On August 25, 2026 Nvidia announced Jetson Orin Nano 2, an entry-level robotics computer with 2x the inference performance of Orin Nano Super, expected in the first half of 2027.
Open-weight 120-billion-parameter mixture-of-experts model that activates 12 billion parameters per token, for text, image, document and video agent work. Nvidia released the weights on Oct. 15, 2026, with Nemotron 3.5 Bolt 74B-A9B for 24GB local systems. Bolt E4B is scheduled for Oct. 23.
Windows laptops built around an NVIDIA Blackwell RTX GPU and Grace CPU, on sale from OEM partners starting at $2,599 as of October 16, 2026. NVIDIA states up to 128GB of unified memory and up to 1 petaflop of FP4 performance. The named lineup includes ASUS ProArt P14 and P16, Dell XPS 16, HP OmniBook Ultra and OmniBook X 14, Lenovo Yoga Pro 9n, Microsoft Surface Laptop Ultra, and MSI Prestige N16 Flip AI+ and Prestige N16. Nvidia does not list which memory sizes shipped.
04
SpaceX
Partially reusable two-stage medium-lift launch vehicle. The world's most frequently launched orbital rocket.
SpaceX
Heavy-lift launch vehicle consisting of a strengthened Falcon 9 core with two additional boosters.
SpaceX
Reusable spacecraft for cargo and crew transport to the International Space Station.
SpaceX
Fully reusable super-heavy-lift launch system designed for missions to the Moon, Mars, and beyond. V3 (Block 3) variant with Raptor 3 engines is the production version.