LLM Reference
activeArena

Chatbot Arena

Metric: Elo Rating (higher is better)Introduced: 2023

Crowdsourced human preference leaderboard using pairwise blind comparisons of LLM responses. Operated by LMSYS/UC Berkeley; rebranded and moved to lmarena.ai in 2024. High benchmark score alone doesn't make a model the right pick — weigh it against pricing, API availability, and release date.

Models ranked

40

tracked on this benchmark

Score band

1503.0 – 1210.0

best → lowest tracked

Snapshot trend

-38.00

May 21 → Jul 3 · 1 models

Leaderboard

Tracked models ranked by Elo Rating (higher is better).

Compare candidates
#Model variant and provenanceScore
1
Claude Opus 4.7
Version: ThinkingHarness: Not recordedEvaluator: Not recordedObserved: Apr 28, 2026Confidence: Not recordedSource
1503.0
2
Claude Opus 4.6
Version: ThinkingHarness: Not recordedEvaluator: Not recordedObserved: Apr 28, 2026Confidence: Not recordedSource
1501.0
3
Gemini 3.1 Pro Preview
Version: Not recordedHarness: Not recordedEvaluator: Not recordedObserved: Apr 28, 2026Confidence: Not recordedSource
1493.0
4
Muse Spark
Version: Arena EloHarness: Not recordedEvaluator: Not recordedObserved: May 5, 2026Confidence: Not recordedSource

Notes: Confidence: high. DAT-3780 weekly best-of refresh: Arena snapshot placed Muse Spark at 1491 Elo.

1491.0
5
GPT-5.5
Version: HighHarness: Not recordedEvaluator: Not recordedObserved: Apr 28, 2026Confidence: Not recordedSource
1488.0
6
Gemini 3 Pro
Version: Not recordedHarness: Not recordedEvaluator: Not recordedObserved: Apr 28, 2026Confidence: Not recordedSource
1486.0
7
GPT-5.4
Version: HighHarness: Not recordedEvaluator: Not recordedObserved: Apr 28, 2026Confidence: Not recordedSource
1479.0
8
ERNIE 5.1
Version: Arena EloHarness: Not recordedEvaluator: Not recordedObserved: May 12, 2026Confidence: Not recordedSource

Notes: Confidence: high. DAT-4172 May 12 /best/ refresh; ERNIE release notes and Arena snapshot placed ERNIE 5.1 around 1476 Elo.

1476.0
9
Qwen3.7-Max
Version: Arena EloHarness: Not recordedEvaluator: Not recordedObserved: May 21, 2026Confidence: Not recordedSource

Notes: DAT-4735 handoff reported Qwen3.7-Max-Preview on Arena AI text leaderboard with scores varying 1475-1489 across sources; using the conservative 1475 value from the datapack.

1475.0
10
GLM-5.1
Version: text (June 2026)Harness: Not recordedEvaluator: Not recordedObserved: Jun 16, 2026Confidence: Not recordedSource
1475.0
11
Gemini 3 Flash
Version: Not recordedHarness: Not recordedEvaluator: Not recordedObserved: Apr 28, 2026Confidence: Not recordedSource
1467.0
12
Claude Opus 4.5
Version: Thinking 32KHarness: Not recordedEvaluator: Not recordedObserved: Apr 28, 2026Confidence: Not recordedSource
1466.0
13
Grok 4.1
Version: ThinkingHarness: Not recordedEvaluator: Not recordedObserved: Apr 28, 2026Confidence: Not recordedSource
1464.0
14
Kimi K2.6
Version: Not recordedHarness: Not recordedEvaluator: Not recordedObserved: May 17, 2026Confidence: Not recordedSource
1462.0
15
Claude Sonnet 4.6
Version: Not recordedHarness: Not recordedEvaluator: Not recordedObserved: Apr 28, 2026Confidence: Not recordedSource
1459.0
16
DeepSeek V4 Pro
Version: text (June 2026)Harness: Not recordedEvaluator: Not recordedObserved: Jun 16, 2026Confidence: Not recordedSource
1456.0
17
DeepSeek V4 Flash
Version: LMArena Chatbot Arena text leaderboard Elo, deepseek-v4-flash-thinkingHarness: Not recordedEvaluator: Not recordedObserved: Jul 3, 2026Confidence: Not recordedSource
1437.0
18
Gemini 3.1 Flash-Lite
Version: Arena EloHarness: Not recordedEvaluator: Not recordedObserved: May 12, 2026Confidence: Not recordedSource

Notes: Confidence: high. DAT-4172 May 12 /best/ refresh; official Google launch post reported Arena Elo 1432.

1432.0
19
o3
Version: Not recordedHarness: Not recordedEvaluator: Not recordedObserved: Apr 15, 2026Confidence: Not recordedSource
1412.0
20
Grok-3
Version: Not recordedHarness: Not recordedEvaluator: Not recordedObserved: Apr 15, 2026Confidence: Not recordedSource
1405.0
21
Gemini 2.5 Pro
Version: Not recordedHarness: Not recordedEvaluator: Not recordedObserved: Apr 15, 2026Confidence: Not recordedSource
1398.0
22
o1 (12-17)
Version: Not recordedHarness: Not recordedEvaluator: Not recordedObserved: Apr 15, 2026Confidence: Not recordedSource
1385.0
23
DeepSeek R1
Version: Not recordedHarness: Not recordedEvaluator: Not recordedObserved: Apr 15, 2026Confidence: Not recordedSource
1372.0
24
Llama 4 Maverick 17B Instruct FP8
Version: Not recordedHarness: Not recordedEvaluator: Not recordedObserved: Apr 15, 2026Confidence: Not recordedSource
1365.0
25
Claude 3.5 Sonnet
Version: Not recordedHarness: Not recordedEvaluator: Not recordedObserved: Apr 15, 2026Confidence: Not recordedSource
1340.0
Showing the top 25 of 40 tracked models. Browse all models.

How to read this benchmark

This benchmark scores models where higher is better. Scores are useful for directional filtering and shortlisting — not for universal quality ranking. Prefer benchmarks closest to your workload, then validate the linked model pages for pricing, context window, and provider availability.

Trust this score when

  • There is a fresh timestamped snapshot (or multiple snapshots) for this benchmark.
  • The model list covers the same version family you can actually deploy today.
  • Top candidates overlap with your required routing and feature requirements.

Be cautious when

  • There is only one benchmark snapshot or the dataset appears stale.
  • The benchmark metric direction is opposite of your decision objective.
  • The score difference between options is narrow and likely within implementation variance.

FAQ

What does the Chatbot Arena benchmark measure?

Crowdsourced human preference leaderboard using pairwise blind comparisons of LLM responses. Operated by LMSYS/UC Berkeley; rebranded and moved to lmarena.ai in 2024. On this page it lists 40 tracked model variants where higher is better.

Is a higher Chatbot Arena score always better?

For this benchmark, higher is better. A high score helps you shortlist, but confirm pricing, context window, and provider availability on each model page before committing — the top scorer is not always the right pick for your workload or budget.

How current is this Chatbot Arena data?

This benchmark was last reviewed on Apr 15, 2026. The tracked score average moved -38.00 points across the last 3 snapshots.

Related benchmarks

Last reviewed: Apr 15, 2026