Abacus

AI Agent Store's weekly roundup for the week of September 14, 2026 corroborates Abacus.AI's Smaug open-weight release, citing the September 10 launch and a more detailed write-up published September 13. The item restates the three-model lineup — Smaug Agentic for long-running coding loops, Smaug Flash for always-on per The roundup adds no new technical specifications beyond the primary Abacus.AI announcement and serves mainly as a timestamp cross-check and a third-party confirmation of the Smaug agentic-workload positioning. Performance and cost claims (15–20% agentic-loop lift, 10–100× cheaper than frontier labs) trace back to Abacu

Shattered.io's September 13, 2026 recap covers Abacus.AI's September 10 release of the Smaug open-weight model family — Agentic, Flash, and Mini — available on Hugging Face and via Abacus.AI's RouteLLM API for teams that don't want to manage inference themselves. The article highlights the open-weight versus open-sourc The piece frames Smaug as a fine-tuning layer over existing open-weight base models rather than a new architecture, citing Abacus.AI's 15–20% lift in long-running agentic-loop performance at unchanged compute cost and the company's positioning against Anthropic and OpenAI subscription pricing. It situates Abacus.AI alo

HPCwire's September 11, 2026 coverage confirms Abacus.AI's release of the Smaug family of open-weight LLMs optimized for enterprise agentic AI, distributed on Hugging Face and designed for hosting inside an enterprise's cloud VPC for data control and privacy. The family consists of Smaug Flash (fine-tuned on DeepSeek F According to the announcement, Smaug's fine-tuning technique improves long-running self-improving agentic loops by 15–20% without raising cost, enabling deployment at open-source price points the company frames as 10–100× lower than frontier models from Anthropic and OpenAI. Security- and privacy-conscious enterprises

Abacus.AI announced the Smaug line of open-weight language models on September 10, 2026, via a PR Newswire release distributed through Yahoo Finance. The launch spans Abacus.AI's enterprise agentic AI platform and Super Assistant, and introduces three models: Smaug Agentic (a 2T-parameter model described as a fine-tune According to Abacus.AI, Smaug is a vendor-developed fine-tuning technique applicable to any open-source base model, which the company claims improves long-running agentic loop performance by 15–20% without added compute cost, framing the release as 10–100× cheaper than frontier models from Anthropic and OpenAI. The sup

On September 10, 2026, Abacus.AI introduced the Smaug line of three open-weight language models — Smaug Agentic, Smaug Flash, and Smaug Mini — fine-tuned for enterprise agentic workloads and released on Hugging Face with availability through the company's RouteLLM API. The launch spans Abacus.AI's enterprise agentic AI Abacus.AI describes Smaug as a fine-tuning technique applicable to any open-source base model, claiming it lifts long-running agentic-loop performance by 15–20% without added cost and positioning the line at 10–100× lower cost than frontier models from Anthropic and OpenAI. CEO Bindu Reddy framed the release as targeti

Abacus AI released Abacus AI Bot, a free, open-source AI agent that runs locally on Windows, Mac, or Linux and connects to messaging and productivity tools without requiring a credit card. The client-side application automates routines, handles multi-step tasks, and supports persistent memory across sessions, per the September 21, 2026 press release. Every new free account receives 2,000 Abacus AI credits, and users can connect their own external model subscriptions via API keys for continued use after credits are exhausted. The release emphasizes user control, persistent context, and a versatile personal-assistant experience spanning chat, research, and background automations.