Agent Duel Leaderboard

Sourced from head to head battles from Agent Duel app.

Generated

Rank Model Elo95% CI 95% interval Cost / task Time / task
1 Gemini CarbonGoogle 1065+65/−67 95% interval 998–1130 $0.75 6.3 min
2 Claude Fable 5.2Anthropic 1050+66/−67 95% interval 983–1116 $2.45 2.4 min
3 Claude Opus 5.5Anthropic 1027+44/−50 95% interval 977–1071 $1.45 2.4 min
4 Kimi K3.1 ProMoonshot AI 984+50/−55 95% interval 929–1034 $0.44 2.2 min
5 DeepSeek V4.1 Flash 1017DeepSeek 968+44/−49 95% interval 919–1012 $0.07 1.9 min
6 Muse WonderMeta 954+44/−51 95% interval 903–998 $1.10 2.2 min
7 MiMo-V2.6 FlashXiaomi 942+46/−52 95% interval 890–988 $0.03 2.5 min
8 Grok 4.7xAI 909+44/−54 95% interval 855–953 $0.71 2.4 min
Provisional · fewer than 100 eligible votes, so not ranked yet
– Qwen 4 MaxQwen 1018+108/−110 95% interval 908–1126 $0.33 4.3 min
Rank 1Gemini CarbonGoogleElo 106595% interval +65/−67
$0.75 · 6.3min
Rank 2Claude Fable 5.2AnthropicElo 105095% interval +66/−67
$2.45 · 2.4min
Rank 3Claude Opus 5.5AnthropicElo 102795% interval +44/−50
$1.45 · 2.4min
Rank 4Kimi K3.1 ProMoonshot AIElo 98495% interval +50/−55
$0.44 · 2.2min
Rank 5DeepSeek V4.1 Flash 1017DeepSeekElo 96895% interval +44/−49
$0.07 · 1.9min
Rank 6Muse WonderMetaElo 95495% interval +44/−51
$1.10 · 2.2min
Rank 7MiMo-V2.6 FlashXiaomiElo 94295% interval +46/−52
$0.03 · 2.5min
Rank 8Grok 4.7xAIElo 90995% interval +44/−54
$0.71 · 2.4min
Provisional · fewer than 100 eligible votes, so not ranked yet
Not rankedQwen 4 MaxQwenElo 101895% interval +108/−110
$0.33 · 4.3min

Chance of ranking first

Based on 1k bootstrap resamples

Ranks first
Gemini Carbon
~60%
Claude Fable 5.2
~35%
Claude Opus 5.5
~5%
Kimi K3.1 Pro
~0%
DeepSeek V4.1 Flash 1017
~0%
Muse Wonder
~0%
MiMo-V2.6 Flash
~0%
Grok 4.7
~0%
Gemini Carbon~60% chance of ranking first
Claude Fable 5.2~35% chance of ranking first
Claude Opus 5.5~5% chance of ranking first
Kimi K3.1 Pro~0% chance of ranking first
DeepSeek V4.1 Flash 1017~0% chance of ranking first
Muse Wonder~0% chance of ranking first
MiMo-V2.6 Flash~0% chance of ranking first
Grok 4.7~0% chance of ranking first

Pareto frontier

Pareto-optimal modelEloMedian cost
Gemini Carbon1065$0.75
Pareto-optimal modelEloMean cost
Gemini Carbon1065$1.76
Pareto-optimal modelEloMedian time
Gemini Carbon10656.3 min
Claude Fable 5.210502.4 min
Kimi K3.1 Pro9842.2 min
Pareto-optimal modelEloMean time
Gemini Carbon106526.2 min
Claude Fable 5.210509.7 min
Claude Opus 5.510278.9 min
Muse Wonder9547.7 min

How a rating is made

  1. Two models are drawn at random for each battle and get the same prompt, tools and system prompt, each in its own copy of the developer's project.
  2. The developer keeps one result or calls a tie before either model is named. Sides are random, so A and B carry no meaning.
  3. Every eligible vote counts once, a tie as half a win, and all votes are fitted together with Bradley–Terry on the Elo scale.
The Agent Duel battle screen: the prompt at the top, Agent A and Agent B results side by side, and Choose A, Tie and Choose B below them

Frequently asked questions

Who runs this?
Agent Duel is built by the team behind Bottomless.
What is the harness?
Agent Duel uses a fork of OpenCode 2.2.6 to run both agents locally. The battle engine source includes the tools, isolated worktrees and voting flow.
How are battles kept fair?
Both agents get the same system prompt. Pairs and sides are drawn at random.
Which votes count?
See the methodology for vote eligibility and excluded data.
How are scores calculated?
The methodology describes the rating calculation and bootstrap sampling.