Another DeepSeek Moment for OpenSource AI?
10th Nov | Superintelligence Newsletter
Hey Superintelligence Fam đ
The AI world is buzzing again - another DeepSeek moment has arrived. Kimi K2 has stunned the global AI race, performing at par with closed-source titans like GPT-5 and Claude 4.5, proving that open-source innovation is far from slowing down.
Meanwhile, the ecosystem is exploding with breakthroughs - from Googleâs Ironwood TPU and Chinaâs chip subsidies to compact marvels like SmolLM2 and ethically charged policy shifts worldwide. Open models are no longer just catching up; theyâre defining the future of intelligence itself.
Read Time : 3 Mins
Why DIY your marketing when you can AI it? Get all your content created for you â instantly.
Running a business is hard - but marketing it doesnât have to be. With Constant Contactâs AI marketing tools, you can create engaging emails, social posts, texts, and full campaigns in minutes, not months. Just type a few words about your business, choose your tone, and let AI handle the rest. Get fresh ideas, ready-to-send content, and faster results - all in one place. Start your free Constant Contact trial today.
Kimi K2 thinking surpasses GPTâ5 & Claude 4.5 : A bold open-source model called Kimi K2 reportedly outpaces GPT-5 and Claude 4.5 in cognitive reasoning and creativity, reshaping the competitive landscape of advanced AI models.
Google launches âIronwoodâ TPU for enhanced AI performance on cloud : Googleâs new Ironwood TPU offers a leap in AI compute power, enabling faster, more efficient large-model training and deployment across its cloud infrastructure.
China powers subsidies for domestic AI chips to challenge global dominance : China is ramping up government subsidies and incentives to boost its domestic AI-chip manufacturing, aiming to reduce reliance on foreign technology and secure strategic sovereignty.
Kimi K2 : A next-generation open-source Mixture-of-Experts model by Moonshot AI with 1 trillion parameters (32 B active), engineered for reasoning, tool-use, and complex multi-agent workflows.
Claude Agent Skills : Anthropicâs Claude Agent Skills lets users create modular âskillsâ â reusable, domain-specific task modules that turn Claude into a specialized expert for workflows, tools, or automation.
Polaris Alpha : Polaris Alpha is a powerful general-purpose AI on OpenRouter featuring a massive 256K-token context window, enabling deep reasoning, coding, and long-context comprehension at scale. (Speculated to be GPT5.1)
SmolLM2: When Smol Goes Big : A 1.7B-parameter model trained on ~11 trillion tokens with rich data-mixing (FineMath, Stack-Edu, SmolTalk) and outperforms models like Qwen2.5-1.5B and Llama3.2-1B.
Global PIQA: Evaluating Physical Commonsense Reasoning Across 100+ Languages and Cultures : A benchmark covering 116 language varieties (5 continents, 14 language families) built by 335 researchers across 65 countries; finds up to 37 % accuracy gaps in physical-commonsense reasoning for lowerâresource languages.
GAP: GraphâBased Agent Planning with Parallel Tool Use and Reinforcement Learning : Introduces a framework where agents build dependency graphs to decide which tools to call in parallel vs sequentially - reducing execution overhead by ~33 % and improving accuracy on multi-hop benchmarks.
UNESCO established pioneering neurotechnology ethics standards to govern brainâmachine research, while Pope Leo XIV urged developers to treat AI creation as a moral responsibility. Legal scholars warned of ethical lapses in AI-assisted filings, while the EU AI Act faced potential delays amid Big Tech lobbying and compliance debates.
Thank you for tuning in to this weekâs edition of Superintelligence Newsletter! Stay connected for more groundbreaking insights and updates on the latest in AI and superintelligence.
For more in-depth articles and expert perspectives, visit our website | Have feedback? Provide feedback.
To explore sponsorship opportunities : Explore Here
Stay curious, stay informed, and keep pushing the boundaries of whatâs possible!
Until Next Time!
Superintelligence Team.