NVIDIA Pushes AI Agents to 3,400 Tokens per Second

Sponsored by

Smarter CRM. Less Busywork.

Disconnected data and tools make it harder to understand your customers. HubSpot's Agentic Customer Platform brings your data, teams, and tech stack together with AI built in to help your business work faster and create more personalized customer experiences.

Why HubSpot and what's new

• Use AI powered tools to take action faster

• Unify your data, teams, and tech stack in one place

• Create one shared view of customer data

• Connect teams around the same customer context

• Bring your business tools into one place

Connect more of your business in one place and give every team a smarter way to work. Get set up quickly and start checking off your hardest tasks.

Get started free

Okta Gives AI Agents Their Own Enterprise Identities

Okta made Agent SSO generally available, bringing the open Cross App Access standard into the identity platform used by 20,000+ customers. Organizations can register agents such as Claude in Universal Directory, apply centralized access policies, monitor activity and change permissions without sharing static credentials.

The strategic shift: AI agents are becoming governed members of the enterprise workforce. Identity providers are emerging as the control layer that determines which systems each agent can access and what it can do.

Britive Eliminates Standing Privileges for AI Agents

Britive introduced Agentic Runtime Control, which verifies an agent and its human delegator, evaluates the task and prompt context, grants only the required access and revokes it when the work ends. Its MCP Gateway can enforce policies at the level of individual tool calls, SSH commands and SQL statements while recording the identities, permissions, actions and outcomes involved.

The strategic shift: agent security is moving beyond protecting credentials toward granting temporary, task-specific authority that can be monitored and withdrawn in real time.

NVIDIA Groq 3 LPX Pushes Agent Inference to 3,400 Tokens per Second

NVIDIA put Groq 3 LPX into full production as an inference accelerator for latency-sensitive agent workloads. It generated a claimed 3,400 output tokens per second on Gemma 4 31B with a 100,000-token context and delivered 4x faster agent responsiveness than the nearest alternative in NVIDIA’s cited benchmarking. Nebius is the first cloud provider adopting it.

The strategic shift: as agents perform hundreds or thousands of reasoning and tool-use steps, token-generation speed is becoming a direct constraint on how quickly they can complete useful work.

SpaceXAI Takes NVIDIA’s Agent CPU From Data Centers Into Orbit

SpaceXAI will deploy NVIDIA Vera CPUs alongside Vera Rubin infrastructure for the next generation of Grok agent workloads as it scales toward gigawatts of computing capacity. The company also plans to use an optimized Vera Rubin NVL72 system in its first-generation Starmind AI satellite. Vera has 88 Olympus cores, up to 1.2 TB/s memory bandwidth and claimed task completion up to 1.8x faster than x86 CPUs.

The strategic shift: agent infrastructure is expanding beyond GPUs. CPU orchestration, code execution, simulations and tool use are becoming important enough to require specialized hardware.

Intel Designs Three Chip Architectures Around Agentic AI

Intel presented an agent-focused hardware stack spanning data centers and edge devices. Diamond Rapids supports up to 256 cores, 16 memory channels and 128 PCIe Gen6 lanes; Crescent Island combines 480GB of LPDDR5X memory with a 350-watt air-cooled design for concurrent agents and long contexts; Wildcat Lake brings up to 17 TOPS of NPU performance to mainstream devices.

The strategic shift: chipmakers are redesigning full hardware portfolios around agent orchestration, inference economics and on-device execution rather than treating agents as conventional model-serving workloads.

Caddi Launches an Agent That Builds Other Agents

Caddi launched a supervisory agent that discovers repetitive business processes, learns each process like a new employee and creates specialized agents to perform the work. It maintains governance and a complete audit trail, initially targeting back-office operations in wealth management, legal services and insurance.

The strategic shift: agent creation itself is becoming an automated business process. Instead of configuring every workflow manually, companies may deploy supervisor agents that identify automation opportunities and generate the required worker agents.

Learn how to code faster with AI in 5 mins a day

You're spending 40 hours a week writing code that AI could do in 10.

While you're grinding through pull requests, 200k+ engineers at OpenAI, Google & Meta are using AI to ship faster.

How?

The Code newsletter teaches them exactly which AI tools to use and how to use them.

Here's what you get:

• AI coding techniques used by top engineers at top companies in just 5 mins a day

• Tools and workflows that cut your coding time in half

• Tech insights that keep you 6 months ahead

Sign up and get access to the Ultimate Claude code guide to ship 5X faster.

Join 200K+ engineers

🗺️ The Map of AI Agents (Live & Growing)

We’re charting the entire AI agent ecosystem — thousands of options across categories.
Your next agent is already on the map.

👉 Explore the Landscape Map →