BenchLeader

LMArena Agent Task Completion leaderboard

As of 16 Nov 2026, the top score on LMArena Agent Task Completion is 22 by Claude Fable 5.2 (Anthropic), ahead of Claude Opus 5.5 at 20.7, across 58 model configurations with a published result (Arena (LMArena), result dated 11 Nov 2026). Share of real agentic sessions (0–100) whose user said the task was completed. LMArena Agent Task Completion is shown for reference only and does not count toward the BenchLeader Index.

Key facts (as of 16 Nov 2026)

  • Claude Fable 5.2 holds the top Arena Agent Task Completion score at 22 (Arena (LMArena), 11 Nov 2026).
  • Claude Opus 5.5 is second on Arena Agent Task Completion at 20.7 (Arena (LMArena), 11 Nov 2026).
  • Arena Agent Task Completion: Share of real agentic sessions (0–100) whose user said the task was completed (Arena (LMArena)).
  • 58 model configurations have a published Arena Agent Task Completion result on BenchLeader (BenchLeader, 16 Nov 2026).
  • Arena Agent Task Completion does not feed the BenchLeader Index; it is shown for reference (BenchLeader methodology).

Share of real agentic sessions (0–100) whose user said the task was completed.

Published by
Arena (LMArena)
Category
Agents & tools
Index weight
Reference only
Models
58
Data as of
16 Nov 2026

CC BY 4.0 — lmarena-ai/leaderboard-dataset on Hugging Face.

58 of 58
#
1NewClaude Fable 5.2max2272.72026-11-11
2Claude Opus 5.5max20.768.42026-11-11
3Claude Sonnet 5.5max19.564.22026-11-11
4Gemini 4 Carbonhigh18.468.92026-11-11
5Claude Fable 5.1max14.466.72026-11-11
6GPT-6 Astramax1467.32026-11-11
7Gemini 4 Argonhigh13.265.32026-11-11
8GPT-6.1 Solmax11.766.62026-11-11
9Muse Wonder Fullmax11.161.72026-11-11
10Claude Opus 5max10.265.02026-11-11
11GPT-6.1 Terramax9.764.12026-11-11
12Hy46.756.72026-11-11
13Muse Wondermax5.460.82026-11-11
14Claude Fable 5high5.163.92026-11-11
15Grok 4.7xhigh5.157.42026-11-11
16MiMo-V2.6-Pro557.52026-11-11
17GPT-6 Solmax4.562.62026-11-11
18Qwen3.8 Maxmax4.558.72026-11-11
19MiMo-V2.6-Flash4.254.82026-11-11
20Muse Spark 1.3max3.760.62026-11-11
21Kimi K3max3.460.22026-11-11
22Inkling 2 Small3.357.92026-11-11
23GLM 5.3max3.158.82026-11-11
24Kimi K3.1 Promax2.759.92026-11-11
25GPT-5.6 Solxhigh2.363.92026-11-11
26DeepSeek V4.1 Flash 1017max2.357.72026-11-11
27GLM 5.2max1.758.92026-11-11
28Claude Haiku 5.5high1.655.32026-11-11
29Gemini 3.8 Flashhigh1.359.42026-11-11
30GLM 5.3 Flash1.357.82026-11-11
31Qwen3.8-Flash-Next0.353.92026-11-11
32GPT-6 Lunamax-0.455.12026-11-11
33Claude Sonnet 5high-0.659.02026-11-11
34MiniMax M3.1 Flash-0.754.32026-11-11
35Step 5 Preview-1.556.02026-11-11
36Grok 4.6xhigh-2.656.12026-11-11
37Qwen3.8 27B-2.755.12026-11-11
38Gemini 3.7 Flashhigh-3.562.12026-11-11
39Mistral Large 4-4.752.92026-11-11
40GPT-6 Lunamedium-6.349.62026-11-11
41Muse Spark 1.1-6.562.22026-11-11
42GPT-5.6 Terraxhigh-7.260.12026-11-11
43NewMuse Embermax-7.652.12026-11-11
44Beam 1.0 501B A23B-9.254.22026-11-11
45Muse Spark 1.2xhigh-9.861.52026-11-11
46Spark X2.6-10.2–2026-11-11
47Qwen3.7max-10.560.02026-11-11
48Qwen3.7 Plus-10.857.62026-11-11
49Gemini 3.1 Pro-11.359.92026-11-11
50Hy3-11.855.72026-11-11
51GPT-6 Lunalow-15.145.82026-11-11
52Mistral Medium 3.5-16.846.62026-11-11
53MiniMax M3-1856.12026-11-11
54Nemotron 3.5 Super VL 120B A12B-18.950.42026-11-11
55Nemotron 3 Ultra-22.842.12026-11-11
56Inkling-26.850.42026-11-11
57Inkling Small-27.652.02026-11-11
58Solar Pro 4-29.948.92026-11-11

AI agent benchmark leaderboard · All LLM benchmarks

Cite as: BenchLeader, “Arena Agent Task Completion leaderboard”, data as of 16 Nov 2026, https://www.benchleader.com/benchmarks/lmarena_agent_task.

Arena Agent Task Completion: questions

What does Arena Agent Task Completion measure?
Share of real agentic sessions (0–100) whose user said the task was completed. Scores are reported in score; higher is better.
What is the top score on LMArena Agent Task Completion?
As of 16 Nov 2026, the highest score on Arena Agent Task Completion is 22 by Claude Fable 5.2 (result dated 11 Nov 2026), ahead of Claude Opus 5.5 at 20.7. Scores come from Arena (LMArena) and are re-read daily.
Which AI model leads Arena Agent Task Completion?
Claude Fable 5.2 leads Arena Agent Task Completion with 22 as of 16 Nov 2026 (result dated 11 Nov 2026), ahead of Claude Opus 5.5 at 20.7.
How many models have Arena Agent Task Completion results?
58 model configurations have a Arena Agent Task Completion result on BenchLeader, all taken from Arena (LMArena).
Who runs Arena Agent Task Completion and how often is it updated?
Arena Agent Task Completion is published by Arena (LMArena). BenchLeader re-reads the published results once a day and records the date each result was published.
Does Arena Agent Task Completion count toward the BenchLeader Index?
No. Arena Agent Task Completion is shown for reference but left out of the composite index.