Benchmark Almanac

Opening Lines Arena

GPT-6 Astra 1,613 leads since September 2026
Not in model cards

Models write opening lines for stories. Pairwise votes determine the published Arena rating.

The official API publishes an Arena rating, confidence interval and vote count. Voter identity requires confirmation from the board description.

Made by
Kaggle
Kind
Creative writing
Level
Everyday
Scored
Ranked by votes
Answers
Open-ended
Tasks
All public
Strongest model tested
GPT-6 Astra #3 on the capability index
Last model added
September 2026
Scores refreshed
November 12th, 2026
15
Models scored on its own leaderboard

GPT-6 Astra leads with 1,613

Best score per model from the benchmark's own leaderboard, read on November 12th, 2026. Where a model ran with several settings or harnesses, we keep its best.

15 models. Scroll sideways for the rest. Bars start at 1,300.

  1. 1,613 GPT-6 Astra
  2. 1,602 GPT-5.6 Terra
  3. 1,596 Grok 4.7
  4. 1,576 GPT-5.6 Sol
  5. 1,567 GPT-5.5
  6. 1,559 GPT-5.2
  7. 1,519 Gemini 4 Argon
  8. 1,515 GPT-6 Luna
  9. 1,455 GLM-5.3
  10. 1,450 Kimi K3
  11. 1,447 Claude Opus 5.5
  12. 1,425 Claude Sonnet 5.5
  13. 1,416 GPT-6.1 Sol
  14. 1,392 Gemini 3.8 Flash
  15. 1,328 GPT-5

Bars start at 1,300.

  1. 1 GPT-6 Astra GPT-6 Astra Max Reasoning 1,613
  2. 2 GPT-5.6 Terra GPT 5.6 Terra Max Reasoning 1,602
  3. 3 Grok 4.7 Grok 4.7 Xhigh Reasoning 1,596
  4. 4 GPT-5.6 Sol GPT-5.6 Sol Max Reasoning 1,576
  5. 5 GPT-5.5 GPT-5.5 High Reasoning 1,567
  6. 6 GPT-5.2 GPT-5.2 Xhigh Reasoning 1,559
  7. 7 Gemini 4 Argon Gemini 4 Argon High Reasoning 1,519
  8. 8 GPT-6 Luna GPT-6 Luna Max Reasoning 1,515
  9. 9 GLM-5.3 GLM-5.3 Max Reasoning 1,455
  10. 10 Kimi K3 Kimi K3 Max Reasoning 1,450
  11. 11 Claude Opus 5.5 Claude Opus 5.5 Max Reasoning 1,447
  12. 12 Claude Sonnet 5.5 Claude Sonnet 5.5 Max Reasoning 1,425
  13. 13 GPT-6.1 Sol GPT-6.1 Sol Max Reasoning 1,416
  14. 14 Gemini 3.8 Flash Gemini 3.8 Flash High Reasoning 1,392
  15. 15 GPT-5 GPT-5 High Reasoning 1,328

Source: Official Opening Lines Arena leaderboard

#ModelScoreSettingDateSource
1 GPT-6 Astra 1,613 GPT-6 Astra Max Reasoning Sep 2026 Official leaderboard
2 GPT-5.6 Terra 1,602 GPT 5.6 Terra Max Reasoning Jul 2026 Official leaderboard
3 Grok 4.7 1,596 Grok 4.7 Xhigh Reasoning Sep 2026 Official leaderboard
4 GPT-5.6 Sol 1,576 GPT-5.6 Sol Max Reasoning Jul 2026 Official leaderboard
5 GPT-5.5 1,567 GPT-5.5 High Reasoning Apr 2026 Official leaderboard
6 GPT-5.2 1,559 GPT-5.2 Xhigh Reasoning Dec 2025 Official leaderboard
7 Gemini 4 Argon 1,519 Gemini 4 Argon High Reasoning Sep 2026 Official leaderboard
8 GPT-6 Luna 1,515 GPT-6 Luna Max Reasoning Sep 2026 Official leaderboard
9 GLM-5.3 1,455 GLM-5.3 Max Reasoning Aug 2026 Official leaderboard
10 Kimi K3 1,450 Kimi K3 Max Reasoning Jul 2026 Official leaderboard
11 Claude Opus 5.5 1,447 Claude Opus 5.5 Max Reasoning Sep 2026 Official leaderboard
12 Claude Sonnet 5.5 1,425 Claude Sonnet 5.5 Max Reasoning Sep 2026 Official leaderboard
13 GPT-6.1 Sol 1,416 GPT-6.1 Sol Max Reasoning Sep 2026 Official leaderboard
14 Gemini 3.8 Flash 1,392 Gemini 3.8 Flash High Reasoning Sep 2026 Official leaderboard
15 GPT-5 1,328 GPT-5 High Reasoning Aug 2025 Official leaderboard

5 models have held number one on the official leaderboard since 2023

Lab claims count from the day the card came out, independent runs from the model’s release.

OpenAI
Own leaderboard 5 leaders
GPT-6 Astra 1,613

Source: Official Opening Lines Arena leaderboard

Scores over time

All the scores we have, one dot each, so a model tested at several settings appears more than once.

On the benchmark’s own leaderboard
1,300 1,400 1,500 1,600 1,700 2026 1,300 1,400 1,500 1,600 1,700 2026

Source: Official Opening Lines Arena leaderboard

No lab model card reports it yet

We read the headline results of 522 launch posts from 52 labs, and none of them list this benchmark.

More in Writing and taste

All 30 in Writing and taste
  • Design Bench (Moonshot) In use Models make graphic designs from direct prompts or few-shot reference images. An automated grader rates the results as Elo scores. Kimi K3.1 Pro 957
  • WritingBench Fading Models complete 1,000 real-world writing requests across 6 domains and 100 subdomains. A judge model scores each reply on five query-specific criteria. GPT-5 83.9
  • Creative Writing (MiniMax in-house) Only in system cards Models write creative pieces for real Hailuo AI user requests. Automatic assessments score the writing. MiniMax-Text-01 81.3%
  • DeepSeek Chinese functional writing evaluation Only in system cards Models answer 3,170 everyday Chinese writing prompts such as reports, emails and speeches. Pairwise judging against a baseline model gives a win rate. DeepSeek-V4-Pro 62.6%
  • Slide generation human evaluation (Z.ai) Only in system cards People compare HTML slide decks from two models on content, layout and visual quality. Metric is overall win rate. GLM-5 67.5%
  • Task-dependent output homogenization evaluation Only in system cards Models answer prompts several times. A judge scores functional diversity where it is wanted and consistency where one answer is right. Muse Spark 0.21