Star Metro - AibleClaw Now Powered by NVIDIA Nemotron 3 Ultra Delivers Frontier-Class Planning for Claws and Enables Stronger Post-Training of Smaller Models

NYSE - LSE

RIO	-0.72%	99.36	$
CMSC	-0.95%	22.16	$
NGG	1.89%	80.97	$
GSK	0.14%	50.74	$
BTI	-0.02%	58.9	$
BP	1.71%	39.78	$
CMSD	-0.95%	22.08	$
RYCEF	1.23%	18.63	$
BCC	-2.92%	72.54	$
BCE	-2.78%	22.65	$
RBGPF	-0.45%	60.34	$
JRI	-0.16%	12.65	$
VOD	-1.27%	14.12	$
AZN	0.85%	176.43	$
RELX	-1.14%	30.83	$

AibleClaw Now Powered by NVIDIA Nemotron 3 Ultra Delivers Frontier-Class Planning for Claws and Enables Stronger Post-Training of Smaller Models

TECHNOLOGY 04.06.2026

AibleClaw Now Powered by NVIDIA Nemotron 3 Ultra Delivers Frontier-Class Planning for Claws and Enables Stronger Post-Training of Smaller Models

Enterprises can now leverage AibleClaw to securely and cost effectively run long-running AI agents or claws with NVIDIA Nemotron 3 Ultra for frontier-class planning and use its outputs to post-train smaller Nemotron 3 Super and Nano models on their enterprise-specific use cases.

Text size:

SAN FRANCISCO, CA / ACCESS Newswire / June 4, 2026 / Aible, the enterprise agentic AI leader, today announced that AibleClaw, its enterprise solution for governed, long-running AI agents, or claws now supports NVIDIA Nemotron 3 Ultra. This newly announced frontier-intelligence open model is smaller, faster, and more cost-efficient than other open models in its class, designed for long-running agents or claws across coding, deep research, and enterprise automation, with up to 5x faster inference and up to 30% lower cost for agentic tasks.

Starting today, Aible customers can now leverage Nemotron 3 Ultra, either by pointing to an existing NVIDIA Cloud Partner endpoint or by having Aible automatically install and configure it on a private server.

Frontier-Class Planning for Claws

In a recent joint hackathon with Aible and the NVIDIA NemoClaw team, AibleClaw was evaluated with NVIDIA Nemotron 3 Ultra against another leading reasoning model using identical OpenClaw configurations running inside NVIDIA OpenShell. The comparison required OpenClaw to find the correct agent across all available agents, identify the right dataset, execute the analysis, post the result to a Slack channel, and save the resulting plan for reliable reuse. This video captures the side-by-side comparison.

Nemotron 3 Ultra planned more directly, executed in less time, and required fewer backtracks than the comparison model. It was the first to post the report to Slack, and the narrative it generated was substantively richer than the comparison model's bullet-point summary with every quantitative claim independently verified by Aible's deterministic hallucination check. Critically, Nemotron 3 Ultra followed all parts of the user instruction on the first try and was able to save the executed plan back as a deterministic NVIDIA AI-Q plan for consistent reuse, while the comparison model missed a specific instruction in the user prompt and failed to correctly call Slack on its first attempt. For enterprise Aible customers, this is the production workflow that matters: a long running AI agent or claw that plans well on the first try, executes reliably, and converts the successful run into a repeatable, deterministic plan that can be scheduled with confidence.

NVIDIA Nemotron 3 Ultra's agentic performance reflects a training approach designed for real-world enterprise workloads, including more than 10 specialized teacher models focused on coding, search, office work, and tool calling, data contributed through the NVIDIA Nemotron Coalition, and multi-turn agent trajectories extracted from agent harnesses.

A Stronger Teacher for Post-Training Smaller Nemotron 3 Models

Aible's "train AI agents like interns, not pets™" approach to post-training, featured in the Nemotron 3 Super announcement, uses fine-grained user feedback on an agent's reasoning steps or tool calling steps to post-train models via NVIDIA NeMo Customizer,NeMo Evaluator, and NeMo Data Designer. A persistent challenge has been the cold-start problem: smaller models often need a larger teacher model to bootstrap the intern-training process. However, the license terms of most frontier models prohibit using their output for post-training.

NVIDIA Nemotron 3 Ultra strengthens this pipeline as the highest-quality teacher model available in the Nemotron 3 family. Because Ultra is part of the same open Nemotron 3 family, with a permissive license, publicly available training data and pipeline, its outputs can be used to post-train smaller Nemotron 3 Super and Nano models. For Aible customers, the result is the same end-to-end post-training pipeline, now with a stronger teacher at the top: bootstrap with Ultra, collect user feedback through the Aible Intern methodology, and post-train the smallest Super or Nano model that meets the customer's accuracy requirements.

To see NVIDIA Nemotron 3 Ultra running with AibleClaw, request a meeting here.

Related Resources

About Aible

Aible is redefining how enterprises drive measurable business impact with specialized and long-running AI agents that are created and coached by business users at scale. Fortune 500 firms, US States, Federal Agencies, and others use AI agents from Aible to deliver measurable results across business areas including customer acquisition and retention, call center optimization, inventory or supply chain management, product development, and risk mitigation in days - not months. Enterprises can also leverage AibleClaw, the secure enterprise solution for long-running AI agents similar to OpenClaw, but with deterministic execution, pre-approved tools, enterprise guardrails, governed data access, and full auditability. The Aible agentic AI platform is the fastest way for business users to automatically look across millions of patterns to surface enterprise insights with cutting-edge augmented analytics, GenAI, and secure long-running agents. Uniquely, Aible gives business users full control of the AI agents by enabling them to provide feedback on the reasoning steps to drive higher accuracy, faster business alignment, and continuously improve performance across thousands of enterprise use cases. Entirely implemented within the security of the enterprise's own cloud or at the edge, the end-to-end serverless AI platform is up to 200X more cost efficient. Learn more at www.aible.com.

Media Contact: [email protected]

SOURCE: Aible

View the original press release on ACCESS Newswire

W.Richard--SMC