# AI News Digest - 2026-10-16

1. Meta released Muse Spark Open, an open-weight model with 816 billion parameters that processed text, images, video, and audio with up to one million tokens of context and was optimized to run locally on Apple's M5 Ultra; it performed roughly on par with leading open-weight Chinese models on benchmarks.

2. Nvidia released Nemotron 3.5 Super VL, an open multimodal model with 120 billion parameters designed to run enterprise AI agents nonstop and accompanied the launch with two compact Bolt models for local deployment; dozens of organizations were already fine-tuning Super VL while Nvidia hardware was noted as advantageous for best performance.

3. Google introduced Gemini 4 Argon as a robotics-focused model that controlled robots, generated simulations from photos or sketches, and evaluated training videos, scoring 87 percent on Google's ERQA benchmark, although published demos still required substantial human guidance.

4. GPT-6-Astra solved 49 of 50 theoretical physics problems on the TheoPhysEval benchmark at a cost of $2.40 per problem, matching specialists' accuracy in the evaluated cases and revealing errors in some experts' reference solutions.

5. Anthropic reported third-quarter revenue of $16.8 billion, a 45 percent year-over-year increase, and disclosed that growth had slowed ahead of a planned November IPO, prompting some investors to reassess its potential valuation.

# References

1. [https://the-decoder.com/muse-spark-open-is-metas-answer-to-qwen-glm-and-kimi-in-the-open-weight-race/](https://the-decoder.com/muse-spark-open-is-metas-answer-to-qwen-glm-and-kimi-in-the-open-weight-race/)

[Muse Spark Open is Meta's answer to Qwen, GLM, and Kimi in the open-weight race](https://the-decoder.com/muse-spark-open-is-metas-answer-to-qwen-glm-and-kimi-in-the-open-weight-race/)

1. [https://the-decoder.com/nvidias-nemotron-3-5-super-vl-is-an-open-model-built-to-run-enterprise-ai-agents-nonstop/](https://the-decoder.com/nvidias-nemotron-3-5-super-vl-is-an-open-model-built-to-run-enterprise-ai-agents-nonstop/)

[Nvidia's Nemotron 3.5 Super VL is an open model built to run enterprise AI agents nonstop](https://the-decoder.com/nvidias-nemotron-3-5-super-vl-is-an-open-model-built-to-run-enterprise-ai-agents-nonstop/)

1. [https://the-decoder.com/google-wants-gemini-4-argon-to-be-the-only-ai-model-robots-need/](https://the-decoder.com/google-wants-gemini-4-argon-to-be-the-only-ai-model-robots-need/)

[Google wants Gemini 4 Argon to be the only AI model robots need](https://the-decoder.com/google-wants-gemini-4-argon-to-be-the-only-ai-model-robots-need/)

1. [https://the-decoder.com/gpt-6-astra-solves-49-of-50-physics-research-problems-at-2-40-apiece/](https://the-decoder.com/gpt-6-astra-solves-49-of-50-physics-research-problems-at-2-40-apiece/)

[GPT-6-Astra solves 49 of 50 physics research problems at $2.40 apiece](https://the-decoder.com/gpt-6-astra-solves-49-of-50-physics-research-problems-at-2-40-apiece/)

1. [https://the-decoder.com/anthropics-q3-revenue-hits-16-8b-but-growth-slows-ahead-of-planned-november-ipo/](https://the-decoder.com/anthropics-q3-revenue-hits-16-8b-but-growth-slows-ahead-of-planned-november-ipo/)

[Anthropic's Q3 revenue hits $16.8B but growth slows ahead of planned November IPO](https://the-decoder.com/anthropics-q3-revenue-hits-16-8b-but-growth-slows-ahead-of-planned-november-ipo/)

For the site tree, see the [root Markdown](https://slashpage.com/ixtj-dev.md).