Benchmarks › LMArena / Arena Text

LMArena / Arena Text

Answer real user prompts in anonymous head-to-head comparisons
1,519top score · Claude Opus 5.5
195models ranked
124open-weights models
Pairwise preference ratinghigher is better

How the test works

1

Prompts

Organic user prompts

  • What the model doesAnswer real user prompts in anonymous head-to-head comparisons
  • Kinds of tasksConversational response generation
  • In → outText → Text
2

Votes

Human pairwise preference votes

  • How runs combineBradley-Terry fit to human pairwise votes, reported on an Elo-like scale
3

Rating

Pairwise preference rating

  • Better isHigher

Arena Text · Creative writing leaderboardtop 10 of 195

Full leaderboard with filters →
ModelArena Text · Creative writingPer taskPriceSpeedSupportsContext
1,519high
$1.82$4 / $20821M
≈1,516high
$1.99$2 / $10841M
1,501high
—$10 / $50—1M
1,500high
—$5 / $25451M
1,489high
—$5 / $25591M
≈1,487high
$1.24$0.75 / $3.751491M
1,485max
$7.63$10 / $50691M
1,485max
$4.52$4 / $201611M
1,481unstated
—$2 / $12—1M
Claude Opus 5Anthropic
1,471max
$5.86$5 / $25551M

LMArena / Arena Text questions

What is LMArena / Arena Text?›

Answer real user prompts in anonymous head-to-head comparisons. Created by LMArena / Arena.

Which model is best on LMArena / Arena Text?›

Claude Opus 5.5 leads Arena Text · Creative writing with 1,519, ahead of Gemini 4 Argon (1,516) and Claude Fable 5 (1,501).

What is the best open-weights model on LMArena / Arena Text?›

Kimi K3 with 1,461, then GLM 5.2 (1,454).

How is LMArena / Arena Text scored?›

Pairwise preference rating; human pairwise preference votes.

Which versions of LMArena / Arena Text are there?›

Arena Text · Creative writing (195 models), Arena Text (196 models), Arena Text · English (188 models), Arena Text · Instruction following (188 models), Arena Text · Non-English (188 models), Arena Text · Russian (186 models), Arena Text · Chinese (181 models), Arena Text · Spanish (155 models).