# Arena Agent net improvement leaderboard

external benchmark

> Every figure below is reproduced as its upstream source published it: nothing is modelled, estimated, interpolated or converted. `Unavailable` means no source published the value — it is never a zero. Each value carries its evidence state and the date it was observed.

Page: https://modelchoice.dev/leaderboards/benchmarks/external-arena-agent-net-improvement

Arena Agent net improvement (last board). Every model the catalog carries a published Arena Agent net improvement value for, ranked by that value.

## Benchmark record

| Field | Value |
| --- | --- |
| Category | external |
| Measure | Net improvement (%) |
| Tasks | Real Agent Mode sessions on arena.ai |
| Difficulty | Multi-step agent work chosen by users |
| Published by | [Arena Agent leaderboard](https://arena.ai/leaderboard/agent) |

Arena's earlier Agent board, which ranked models by net improvement in real Agent Mode sessions, with confirmed success, praise vs complaint, steerability, bash recovery and tool hallucination beside it. Replaced by the Agent Capability Index on November 18, 2026.

## Arena Agent net improvement ranking

No model in the catalog has a published Arena Agent net improvement score. The benchmark is defined by Arena Agent leaderboard, but the catalog carries no value for it yet. An absent value is shown as absent here rather than as a zero.

Rows are ordered by the value the source published, highest first. The source does not state whether a higher value is the better result, so this document does not either: for a benchmark that measures a rate of failure, read the table from the bottom.
