Moonshot AI Launches Kimi K3.1 Pro, 2.8T Parameters

Key Takeaways

• Moonshot AI launched Kimi K3.1 Pro with 2.8T parameters and 1M token context, claiming parity with OpenAI and Anthropic top models on coding and agentic benchmarks.
• Kimi K3.1 Pro scored 55 on the Artificial Analysis Intelligence Index, ranking 4th among proprietary models and gaining 11 points over its predecessor.
• Nokia's AI and Cloud division revenue grew 123% year over year to 631 million euros in Q3 2026, driven by hyperscaler data center demand.
• Nvidia invested $1 billion in Nokia last year, acquiring 2.9% of the company as the AI networking market heads toward $200 billion by 2030.
• Anthropic and Hg are forming a new consultancy with hubs in London and Berlin to embed AI engineers in European enterprises by end of 2026.
• BioStack raised $9.5 million in seed funding to build data infrastructure connecting fragmented clinical information across 100M+ de-identified patient records.
• The South Mississippi Fair will deploy Evolv AI weapons detection when it opens October 23, following a shooting at the fairgrounds in 2024.
• Quad9 will stop serving French users on November 24 rather than comply with a court order to block beIN domains, citing its privacy mission.
• A new study uses Petri nets to keep AI agents safe during tool execution, blocking harmful requests even under prompt injection attacks.
• Nvidia's stock has risen 14% this year as investors reassess AI spending and favor the chipmaker's stability over flashier AI plays.

Moonshot AI launches Kimi K3.1 Pro matching top US models on some benchmarks

Moonshot AI released Kimi K3.1 Pro, a 2.8 trillion parameter mixture-of-experts model with 104 billion active parameters and 1 million token context. The company says it matches or beats OpenAI GPT-6 Astra and Anthropic Opus 5.5 on coding, agentic and knowledge benchmarks. A faster Highspeed variant runs 2 to 3 times quicker. Open weights are promised by November 4, 2026.

Kimi K3.1 Pro ranks 4th on Artificial Analysis Intelligence Index

Kimi K3.1 Pro scored 55 on the Artificial Analysis Intelligence Index, placing 4th behind leading proprietary models. It gained 11 points over its predecessor Kimi K3 despite identical size. Big gains came in agentic tasks, including Terminal-Bench 4.0 where it jumped from 13% to 67.2%. Hallucination rates also fell from 53% to 44% on the AA-Omniscience Index.

Moonshot AI releases Kimi K3.1 Pro with 2.8 trillion parameters and 1 million context

Moonshot AI launched Kimi K3.1 Pro, a model with 2.8 trillion total parameters and 104 billion active parameters using Kimi Delta Attention. It is available on Kimi API with open weights expected in November 2026. On Terminal-Bench 4.0 it scored 65.7%, just behind Claude Opus 5.5, and leads SWE-Marathon at 51.7%. API pricing is $2 input and $10 output per million tokens.

Nvidia-backed Nokia grows AI revenue by 123% through networking infrastructure

Nokia is transforming from a phone company into an AI networking equipment supplier. Nvidia invested $1 billion in Nokia last year, acquiring 2.9% of the company. In Q3 2026, Nokia's AI and Cloud division revenue grew 123% year over year to 631 million euros. The company secured about 1.5 billion euros in new orders during the quarter.

Nokia's AI networking business drives 123% revenue growth as hyperscalers build data centers

Nokia is benefiting from the AI infrastructure build-out by supplying high-speed optical networking equipment. Nvidia invested $1 billion in Nokia last October. Nokia's third quarter revenue reached 5.2 billion euros with operating profit up 35%. The AI-RAN market is expected to exceed $200 billion by 2030 according to Omdia data.

South Mississippi Fair adds AI weapons detection

The South Mississippi Fair in Laurel opens Oct. 23 with a new AI-powered weapons detection system called Evolv. The system scans visitors as they walk through without stopping, and flags possible weapons on a screen. This comes after a teenager was shot at the fairgrounds in 2024. Jones County Sheriff Joe Berlin said extra security officers will patrol the fairgrounds, which run through Oct. 31.

New study addresses safe tool execution for AI agents

A new paper studies how to keep large language model agents safe during tool execution, especially under prompt injection attacks. The researchers use Petri nets to formalize the control system between the AI and its tools. Their approach blocks harmful requests even if they are worded cleverly. The study also introduces a replanning closed loop concept to handle retry attempts and identifies conditions for the least intrusive safe supervisor.

Quad9 will cut France off rather than lie in DNS

Quad9, a Swiss nonprofit DNS resolver, will stop serving French users on Nov. 24 rather than comply with a court order to block beIN domains. The company says it has no user database and cannot identify French queries without compromising its privacy mission. A German case from 2021 to 2023 showed GeoIP blocking is unreliable, as users bypassed it with VPNs. About 20 similar lawsuits are pending against Quad9 in France, making compliance unsustainable for the donor-funded service.

AI agent security is a continuous lifecycle

A new article argues that AI agent security must be a continuous lifecycle, not a one-time checkpoint. AI agents can use inherited human credentials to take actions their users never intended. The article calls for action-level authorization that evaluates what an agent is trying to do before it executes. It gives an example where a coding agent is denied permission to delete a production repository. The piece stresses that robust security requires enforcement at multiple layers, including APIs and tool servers.

Anthropic and Hg form company to embed AI engineers in Europe

Anthropic and the investment firm Hg have formed a new consultancy expected to launch by the end of 2026. The company will have hubs in London and Berlin to embed AI engineers in European enterprises. Hg announced in September that it extended its collaboration with Anthropic to support agentic AI. The partnership aims to help Claude reach large enterprises through systems integration.

Numerix shows AI agent for autocallable pricing

Numerix demonstrated an AI agent that can generate pricing code for autocallable products. The demo used existing documentation and Python examples to create templates for a worst-of barrier reverse convertible. Ping Sun, SVP at Numerix, explained the history of autocallables, noting BNP Paribas traded the first product in the US about 23 years ago. He also stated that autocallables now make up about 40% of the total exotics market.

BioStack raises $9.5 million for medical AI data platform

BioStack raised $9.5 million in seed funding led by Peak XV and Afore Capital. The California startup builds data infrastructure that connects fragmented clinical information across healthcare systems. Its platform creates longitudinal patient records from notes, labs, imaging, medications, and outcomes across 100M+ de-identified patient records in the US, India, Brazil, and Argentina. The funding will support expansion, technology development, and deeper hospital partnerships.

Nvidia stock may offer stability as Wall Street rethinks AI spending

Nvidia's stock has risen 14% this year, matching the S&P 500's gains. Analysts say Nvidia's dependability could now appeal more to investors who previously favored flashier AI plays. Wall Street is taking a more cautious approach to the AI trade, examining the debt behind the AI boom and whether it could derail long-term growth projections. The shift comes as investors reconsider the spending cycle on chips and power.

Sources

• Moonshot AI Releases Kimi K3.1 Pro, Beats Opus 5.5, GPT Astra 6 On Some Benchmarks
• Kimi K3.1 Pro Places 4th On Artificial Analysis Intelligence Index, Ahead Of GPT-6 Astra, Gemini 4 Argon
• Moonshot AI Releases Kimi K3.1 Pro: A 2.8T Total/104B Active Params (MoE) Model with 1M Context
• Not Intel. Not CoreWeave. This Underappreciated Nvidia-Backed Company Is Growing Artificial Intelligence (AI) Revenue by 123%. Here's Why It's a Buy Today for $8.
• Not Intel. Not CoreWeave. This Underappreciated Nvidia-Backed Company Is Growing Artificial Intelligence (AI) Revenue by 123%. Here's Why It's a Buy Today for $8.
• South Mississippi in Fair adds AI weapons detection system
• Certified Execution Harness for Agentic AI: Safe and Least-Intrusive Petri-Net Supervisory Control Between Reasoning and Action
• Quad9 will cut France off rather than lie in DNS
• AI agent security is a lifecycle
• Anthropic and Hg Form Company to Embed AI Engineers in European Enterprises
• Numerix demos agentic AI for autocallable pricing
• BioStack Raises $9.5 Mn to Build Medical AI Training Infrastructure
• Can Nvidia’s stock become a safe haven as Wall Street rethinks the AI trade?