When @Felicis first invested in @Arena at the seed, we believed the company was building something essential: an independent, community-driven source of truth for measuring the AI frontier. That conviction only grew, and we went on to lead the Series A. The progress since then has been extraordinary. Arena has surpassed $100 million in annualized revenue, facilitated 350 million sessions and 62 million votes, and grown into one of the world’s largest AI communities. Today, Arena announced a $200 million Series B at a $3.1 billion valuation, alongside a new Alignment Index that measures whether AI agents behave safely, truthfully, and within the bounds of what users actually ask them to do. As AI systems begin taking real actions on our behalf, measuring capability is no longer enough. We also need to know whether these systems can be trusted. Arena is uniquely positioned to give the industry that answer, grounded in millions of real-world human interactions. We are true believers in @ml_angelopoulos, @infwinston, and the entire Arena team as they build the trusted measurement layer for AI.

Introducing the Arena Alignment Index, our new benchmark measuring safety and alignment of AI agents in real-world use. Built from 90K+ real-world agent sessions across 27 models, the index measures three critical signals: - Unauthorized Action (UA): Taking actions beyond the user's instructions or permissions - False Attribution (FA): Attributing statements or actions that are contradicted by user-provided evidence - Deceptive Completion (DC): Claiming a task was completed when it was not. Key findings: - OpenAI models currently lead the Alignment Index - Rogue actions are rare, but can have serious consequences when they occur - Agents can mislead users about task progress - Misalignment risks increase with conversation length - Safety and alignment are improving across model generations As shown in the leaderboard below (sorted by lab), @OpenAI’s GPT-6.1-Sol leads the Arena Alignment Index with a score of 87.9, followed by @AnthropicAI’s Claude-Opus-5.5 at 83.2 and @SpaceXAI's Grok-4.7 at 82.7. OpenAI also has the best observed rates across all three signals: 0.89% Unauthorized Action, 1.98% False Attribution, and 2.34% Deceptive Completion. Across all four labs, newer models consistently outperform their predecessors, suggesting broad progress in agent safety and alignment. As agents take on longer, more complex, and higher-stakes tasks, measuring not just what they can accomplish, but how safely and reliably they act, becomes increasingly important. This marks an important step toward making safety and alignment a core part of how Arena evaluates AI. The index is an initial starting point, and we'll continue expanding the index with additional safety signals and models over time. More analysis below👇

We're excited to announce a major milestone for Arena: our $200M Series B at a $3.1B valuation, accelerating our mission to measure and advance frontier intelligence! At Arena, we believe AI evaluation must be grounded in real-world use. We put agents in live tool-use environments, challenge them with complex, long-horizon tasks, and analyze rich signals from millions of real-world user-agent interactions to understand how they perform. Today, we're also introducing the Arena Alignment Index, extending our evaluations beyond capability to measure how safely, reliably, and faithfully agents act in accordance with human intent. As AI systems become increasingly autonomous, evaluating what they can do is no longer enough. Understanding whether they reliably follow human intent, navigate uncertainty, and act safely in real-world environments is becoming one of the most important challenges in AI. We're excited to continue advancing the research and building the infrastructure needed to rigorously measure both capability and alignment at the frontier of AI. If you’re passionate about this mission, come build with us!

Announcing the Fall 2026 cycle of @arena's Academic Partnerships Program. We're accepting new proposals from tenure-track faculty at U.S. universities working on the scientific foundations of AI evaluation: evaluation and ranking methodology, measurement and validity, learning from human preference, and safety and alignment. Funding of up to $50,000 per project is available, and we'll continue reviewing submissions twice a year, each Spring and Fall. Congratulations to our first Spring 2026 cohort of 7 selected research projects! Head to the thread to see who was selected, learn more about the program, and find out how to submit your own proposal for Fall.

Yesterday, my OpenAI collaborator and I gave a detailed talk on the Huggingface incident, our models creating "the message board", model misalignment, and more. I hope it can answer a lot of the questions folks have, and we will release a full detailed postmortem at a later time!

DeepSeek-V4-Flash-20260731 (High) by @deepseek_ai landed in Agent Arena at #21# overall (+1.98% net-improvement), and among open source #3# based on +12.5K real-world agentic sessions. This release is ranked 6 higher than DeepSeek-V4-Pro (-0.47%) and 13 higher than the previous DeepSeek-V4-Flash model (-2.81%). Across key signals, DeepSeek-V4-Flash-20260731 (High) is strong in Confirmed Success (+6.99%), but weaker in Steerability (-2.31%). Below we break down how it scored across all 5 signals, drawn from tasks submitted by a global community of users. Here’s an overview on the signals: - Confirmed Success: an explicit "yes that worked" feedback from the user - Praise vs. Complaint: implicit sentiment in users reactions - Steerability: can the model course-correct when you push back? - Bash Recovery: how it recovers from CLI errors (primary signal for tool use) - Tool Hallucination: does it call tools that don't exist

The latest DeepSeek release is underhyped. DeepSeek-V4-Flash pushes the intelligence/$ frontier forward by an order of magnitude, and it's open-source. This will bring powerful agentic capabilities to the mass market and unlock 10x more use cases overnight.