The day in AI, in five minutes.
Moondream 3 Preview is listed on OpenRouter, while NVIDIA released Nemotron 3.5 Super VL and two Bolt edge models; Black Forest Labs and Xiaomi reportedly released open-weight models. On the Vikshy Score, Meta moves from #5 to #4, Alibaba from #4 to #5, xAI from #7 to #6, Z.ai from #8 to #7, DeepSeek from #6 to #8, Moonshot from #11 to #10 and Tencent from #10 to #11; OpenAI services have recovered, and Moonshot and MiniMax status pages are back to normal.
What mattered
- 1
Moondream 3 Preview vision-language model listed on OpenRouter
Why it matters: Moondream 3 Preview brings visual question answering, captioning, OCR and visual reasoning to OpenRouter, with 9B total parameters and 2B active in a mixture-of-experts design.
Moondream 3 Preview, a vision-language model from Moondream, is listed on OpenRouter. It has 9B total parameters with 2B active in a mixture-of-experts design and handles visual question answering, captioning, OCR and visual reasoning.
- 2
NVIDIA releases Nemotron 3.5 Super VL and Bolt edge models
Why it matters: NVIDIA's open 120-billion-parameter Nemotron 3.5 Super VL and two smaller Bolt models give builders options for always-on agents and local agentic workloads.
NVIDIA released Nemotron 3.5 Super VL, an open 120-billion-parameter model for always-on agents. It also released two edge models, Nemotron 3.5 Bolt 74B-A9B 24GB and Nemotron 3.5 Bolt E4B 4GB, for local agentic workloads.
- 3
OpenAI services recover after widespread feature degradation
Why it matters: OpenAI services have fully recovered after a reported major incident caused widespread feature degradation.
OpenAI reported a major incident causing widespread feature degradation across its services. All impacted services have now fully recovered.
- 4
Black Forest Labs releases open-weights FLUX 3 Image Klein model
Why it matters: According to reports, FLUX 3 Image Klein supports native 4K generation, pixel-precise editing and up to ten reference images, and runs on GPUs with 8 GB of VRAM.
Black Forest Labs released FLUX 3 Image Klein, a 5.6-billion-parameter open-weights image model supporting native 4K generation, pixel-precise editing and up to ten reference images. It has 38 percent fewer parameters than its predecessor and runs on GPUs with 8 GB of VRAM.
- 5
Xiaomi open-sources MiDashengLM-Spatial audio model
Why it matters: According to reports, Xiaomi's Apache 2.0 MiDashengLM-Spatial combines general audio understanding with sound direction, distance and movement.
Xiaomi released MiDashengLM-Spatial, an Apache 2.0 audio model built on MiDashengLM-7B with a new Spatial-Dasheng encoder. It combines general audio understanding with sound direction, distance and movement.
The Vikshy Score
The close of 16 Oct, 06:00 UTC, against the previous close. The full board.
The numbers
Read straight from the source. Every signal.