WorkBench Revisited: Workplace Agents Two Years On
Quick Answer
This paper shows that In June 2026, Claude Opus 4.8 outperformed GPT-4 by completing 89% of tasks with only 2.5% unintended harmful actions.
Quick Take
In June 2026, Claude Opus 4.8 outperformed GPT-4 by completing 89% of tasks with only 2.5% unintended harmful actions. The study reveals that capability and safety are positively correlated, with open-weight models reducing costs significantly while maintaining performance. An updated benchmark with improved data and analysis has been released.
Key Points
- Claude Opus 4.8 achieves 89% task completion, a significant improvement over GPT-4's 43%.
- Unintended harmful actions decreased from 26% with GPT-4 to 2.5% with Claude Opus 4.8.
- Capability and safety improvements are positively correlated in agent performance.
- Open-weight models have drastically reduced costs for high-performance agents.
- An updated benchmark with new data and analysis has been released for further research.
Paper Resources
Article Excerpt
From source RSS / original summaryarXiv:2606. 13715v1 Announce Type: new Abstract: The best agent on WorkBench in March 2024, GPT-4, completed 43% of tasks and took an unintended harmful action, such as emailing the wrong person, on 26% of them. We re-visit the benchmark in June 2026 and find that the best agent to date, Claude Opus 4. 8, completes 89% and takes an unintended harmful action on 2. 5%. Aside from this considerable progress in frontier agent performance, three things stand out.
First, capability and safety go together on WorkBench rather than trade off, so the models that finish the most tasks also do the least unintended damage. Second, while several classes of error have been totally eliminated, frontier models still make some basic mistakes that occasionally result in irreversible harm, such as sending an email to the wrong person.
Third, the rise of open-weight models has drastically lowered costs for a performance level that was previously only accessible to proprietary models, while frontier costs have stayed relatively stable. We release an updated version of the benchmark with data and code quality improvements, new model scores, and analysis of agent progress on WorkBench since 2024.
Reader Mode unavailable (could not extract clean content).
Want this in your inbox every morning?
Daily brief at your local 8am — bilingual EN/中文, free.
More from arXiv cs.AI
See more →Arbor: Tree Search as a Cognition Layer for Autonomous Agents
Arbor introduces a multi-agent framework utilizing structured tree search for optimizing LLM inference, achieving up to 193% throughput-latency improvement compared to vendor-optimized systems. It employs an Orchestrator and Critic agent for stability and coordination, demonstrating hardware-agnostic performance with minimal variance.