ResearchArenaConfirmed
Story 16 of 18 in the Wednesday 18 November editionArena launches Agent Arena Index and six sub-indexes over 2.9M sessions
Updated on 18 November 2026 (methodology posted 16 November) with 1000-anchored longitudinal score scaling across Coding, Work, Chat and Tool Use.
Arena launched the Agent Arena Index and six focused sub-indexes - Agentic Work, Agentic Coding, Agentic Chat, Task Outcome, Steering and Tool Use - alongside its Alignment Index across more than 2.9 million user sessions (methodology published 16 November and updated 18 November 2026), introducing 1000-anchored longitudinal score rescaling while retiring Tool Hallucination and renaming Confirmed Success to Task Completion.
Study details
- Session volume and baseline
- Over 2.9M user sessions; linear 1000-point July anchor (1 point = 1/1000th of a signal SD)
- Seven published indexes
- Agent Arena Index plus Agentic Work, Agentic Coding, Agentic Chat, Task Outcome, Steering and Tool Use
- Signal set updates
- Tool Hallucination retired; Confirmed Success renamed to Task Completion; Alignment Index retained
Release details & context
Linear anchor offset keeps longitudinal scores comparable
Previously, Arena’s inverse-probability-weighted net improvement estimator compared each model against the uniform mean of active rivals, causing an unchanged model’s score to fall whenever stronger entrants arrived. Multiplying by 1,000 and recomputing a constant anchor offset across converged carryover models stabilises historical comparisons without altering rank order.
Sources
- PrimaryArena Blog: Introducing the Agent Arena Indexarena.ai
- On XArena (@arena) Agent Arena Index post on Xx.com
Arena release timeline (2 launches tracked)
- LMSYS Arena adds live tool Connectors to Agent Mode
LMSYS Arena added live tool Connectors to Agent Mode, enabling evaluation of agentic models on real developer infrastructure.


