AI Agents
Apodex 1.1 Scales Agentic Intelligence for Long-Horizon Professional Tasks
The August 24, 2026 release centers on environment scaling and agentic coordination to support sustained verifiable work across finance, research, coding, and math with open harness tools and a 35B local model option.
Apodex 1.1 is a model family released by Apodex on August 24, 2026, that scales agentic intelligence for complex work via environment scaling for executable environments and agentic coordination scaling for long-horizon task decomposition.
Apodex released the 1.1 model family on August 24, 2026.
The release includes the full frontier model for online workbench use.
The release includes the open-weight 35B-parameter Apodex 1.1 Mini for local deployment.
What context explains the need for agentic scaling in models like Apodex 1.1?
General-purpose language models can reason and synthesize knowledge, but complex work also requires sustained interaction with files, information sources, and executable code, together with state maintenance, failure recovery, and verifiable delivery.
Apodex 1.1 addresses these requirements through dedicated scaling approaches.
What new elements appear in the Apodex 1.1 model family release?
Apodex 1.1 emphasizes Environment Scaling for executable file, search, and code environments.
Apodex 1.1 emphasizes Agentic Coordination Scaling for long-horizon task decomposition and replanning.
The release includes open-source FrontierAgent harness and asynchronous Agent Team capabilities.
Bare-model API access is planned.
How does Apodex 1.1 perform on key professional benchmarks?
Apodex 1.1 reaches the leading performance band across professional work, finance, scientific research, mathematics, coding, and search despite smaller model size than many frontier systems.
| Benchmark | Score with Agent Team | Attributed Source |
|---|---|---|
| GDPVal professional work benchmark | 78.8 | arXiv: https://arxiv.org/abs/2608.23283 |
| APEX-Agents professional work benchmark | 38.5 | arXiv: https://arxiv.org/abs/2608.23283 |
| FrontierFinance benchmark | 54.3 | arXiv: https://arxiv.org/abs/2608.23283 |
What capabilities does the 35B Apodex 1.1 Mini provide?
The 35B Apodex 1.1 Mini retains strong working capability for local deployment.
What open-source tools support the use of Apodex 1.1?
Researchers can use files, search, code execution, and Agent Team inside one long-running task.
This turns reasoning into inspectable and reusable work.
- Download the FrontierAgent harness from the official GitHub repository.
- Integrate Agent Team for asynchronous coordination in tasks.
- Deploy the Apodex 1.1 Mini on local hardware for data-sensitive applications.
- Access the full frontier model through the online workbench for complex workflows.
What implications does Apodex 1.1 have for enterprise AI adoption?
The model family supports verifiable progress toward real-world objectives in finance and research.
Local deployment of the Mini model aids in maintaining control over proprietary data.
How do the Apodex statements describe the goals of the 1.1 release?
We’re excited to introduce Apodex 1.1, our new model family built to scale agentic intelligence for professional work.Apodex
We call this working capability: sustained, verifiable progress toward a real-world objective.Apodex Team, Authors
What next steps are outlined for Apodex 1.1 access?
Bare-model API access is planned as part of the ongoing development.
The emphasis on state maintenance allows for continued task progress over extended periods.
Failure recovery mechanisms help maintain progress in the face of errors during execution.
Verifiable delivery ensures that outcomes can be inspected and validated by users.
The online workbench integrates multiple tools for seamless long-running tasks.
Finance professionals can apply the model to complex calculations and data analysis.
Scientific researchers benefit from the coding and search capabilities in one environment.
Mathematics tasks see improved handling through the coordination scaling.
Coding projects gain from the executable code environments provided.
Search functions are enhanced within the agentic setup.
The smaller size of the Mini model facilitates broader accessibility for local use.
Open weights allow for customization by developers and organizations.
The harness provides a standardized way to implement agent teams.
Asynchronous capabilities enable parallel processing of subtasks.
Task decomposition breaks down complex objectives into manageable parts.
Replanning occurs dynamically based on intermediate results.
The overall approach moves beyond simple reasoning to actionable work output.
Frequently asked
When was Apodex 1.1 released?
Apodex released the 1.1 model family on August 24, 2026.
What is the parameter count of Apodex 1.1 Mini?
The Apodex 1.1 Mini has 35 billion parameters and supports local deployment.
Which benchmarks does Apodex 1.1 lead according to available data?
Apodex 1.1 leads in GDPVal with 78.8, APEX-Agents with 38.5, and FrontierFinance with 54.3 when using Agent Team.
Sources
- arXiv — Apodex 1.1 with Agent Team achieves 78.8 win rate on GDPVal professional work benchmark
- Apodex — This release is centered on the Apodex 1.1 model and the online workbench it already powers. Researchers can use files, search, code execution, and Agent Team inside one long-running task, turning reasoning into inspectable and reusable work.
- ApodexAI — The Apodex results are summarized here: Apodex-1.1 Agent Team | 38.5 | 78.8 | 54.3 | 63.3 | 35.3 | 56.1
- Hugging Face — We’re excited to introduce Apodex 1.1, our new model family built to scale agentic intelligence for professional work.