New family of LLMs that gives Enterprises total control of their data, enhanced privacy, and SOTA-level agentic performance

SAN FRANCISCO, Sept. 10, 2026 /PRNewswire/ — Abacus.AI is introducing three new LLMs across its enterprise agentic AI platform and Super Assistant. The model family called Smaug optimizes open-source base models for long-running self-improving agentic loops in a cost-efficient manner. These models can easily be hosted by any enterprise within its cloud VPC environment, giving it full control over its data and where its AI is hosted.

Smaug is a novel fine-tuning technique applicable to any open-source base model that improves the performance of its long-running agentic loops by 15-20% without increasing cost. This fundamentally enables companies to deploy self-improving enterprise AI agents at scale at open-source model costs, typically 10–100× lower than those of frontier models from Anthropic and OpenAI.

Today, we are releasing Smaug Agentic, Smaug Flash, and Smaug Mini.

All of these models will be available on Hugging Face. They are open-weight and can be downloaded by anyone.

Smaug Flash – This model is highly optimized for personal agents and can connect to messaging apps, including WhatsApp, Telegram, and Slack, and maintain long-running conversations with the user, adapting to their vibes over time. This model has been fine-tuned on DeepSeek Flash.

Smaug Agentic – This model achieves state-of-the-art performance on complex coding loops, is based on Kimi K3, and is a large 2T model. This is an excellent replacement for Opus-class models and is more cost-efficient. Enterprises that are concerned about security and data privacy can host Smaug Agentic on an in-house GPU cluster.

Smaug Mini – Smaug Mini is the efficient option in the line and is a smaller 27B model, optimized for multimodal use cases and smaller reasoning workloads. It delivers strong agentic gains in a compact, cost-effective package. This is an excellent option for simple tasks, great for enterprise chatbots, and can be further fine-tuned on the enterprise’s data.

We have also published these models on the popular benchmark LiveBench AI. The comparison to the base models on key categories like agentic coding is available here.

These models deliver meaningful performance gains over their base models across coding, tool use, automation, and long-context reasoning workloads. For complete benchmark results and technical details, read the full technical brief.

Together, the Smaug line of models provides enterprises with a cost-effective option to control their data and automate workloads.

“Open-weight models are rapidly closing the gap to frontier closed models, but still underperform in long-running agent loops. The Smaug line addresses this shortcoming while still being 10-100x cheaper than closed-source models,” said Bindu Reddy, CEO of Abacus.AI.

The Smaug line is both a product release and a demonstration of a broader thesis: with the right fine-tuning methodology, open-weight models can compete with and surpass frontier models on Agentic AI tasks. Abacus.AI expects to continue advancing the line as base models evolve and its library of agentic traces grows.

About Abacus.AI
Abacus.AI is an Enterprise SuperAssistant and agentic platform to automate white-collar work. The platform applies 100+ AI models to create self-improving AI workflows, AI-assisted apps, and an enterprise-wide AI brain and is used by over 3M professionals, small businesses, and enterprises.

Headquartered in San Francisco, Abacus.AI is backed by world-class investors, including Coatue, Tiger Global, and Index Ventures. Abacus.AI was started by two Google, Amazon, and Uber veterans, Bindu Reddy (bindu@abacus.ai) and Arvind Sundararajan (arvind@abacus.ai).

More information here: http://abacus.ai/about

Press inquiries: kim@abacus.ai

Additional graphics, photos, and other assets are available upon request.

View original content:https://www.prnewswire.com/news-releases/abacusai-launches-the-smaug-line-of-open-weight-models-optimized-for-enterprise-agentic-ai-use-cases-302875524.html

SOURCE Abacus.AI