Benchmark Almanac

NARRA-Gym

Claude Sonnet 4.6 2 leads since May 2026
Not in model cards

Agents build interactive stories around a user’s experiences and adapt scenes as the conversation unfolds. Judges rate story quality and user experience separately.

The project leaderboard lists eight models. The paper compares nine models across eight personas using three judges.

Made by
Yue Huang and colleagues
Released
May 2026
Kind
Creative writing
Level
Everyday
Scored
Graded by a model
Answers
Open-ended
Tasks
All public
Measures
Capability
Language
English
In and out
Text → Text, Actions
Interaction
Agentic
Environment
A simulator
Items from
Generated
Refresh
One fixed set
Run by
Anyone
In views
Agents, Robotics
Strongest model tested
GPT-5.4 #28 on the capability index
Last model added
March 2026
Scores refreshed
November 20th, 2026
8
Entries on its own leaderboard, 7 models we track

Claude Sonnet 4.6 leads with 2 among models released by March 2026 Claude Sonnet 4.6 leads with 2 among models released by March 2026

Best score per entry from the benchmark's own leaderboard, read on November 20th, 2026. Where a model ran with several settings or harnesses, we keep its best. Names in gray aren't in our model list: fine-tunes, agent systems, products and older models. They don't count toward leaders. This view shows only models we track, each at its best.

The newest model with a score here came out in March 2026, and nobody has tested a newer one. Among the 20 highest-ranked models on the capability index, these have no score yet:

Bars start at 0.5.

  1. 1 Claude Sonnet 4.6 2
  2. 2 Claude Opus 4.6 2
  3. 3 GPT-5.4 2
  4. 4 Qwen3.5 397B-A17B 1
  5. 5 Gemini 3.1 Pro 1
  6. 6 Doubao 1
  7. 7 DeepSeek-V3.2 1
  8. 8 GLM-5 1

Source: Official NARRA-Gym leaderboard

#EntryScoreSettingDateSource
1 Claude Sonnet 4.6 2 Feb 2026 Official leaderboard
2 Claude Opus 4.6 2 Feb 2026 Official leaderboard
3 GPT-5.4 2 Mar 2026 Official leaderboard
4 Qwen3.5 397B-A17B 1 Feb 2026 Official leaderboard
5 Gemini 3.1 Pro 1 Feb 2026 Official leaderboard
6 Doubao 1 Official leaderboard
7 DeepSeek-V3.2 1 Dec 2025 Official leaderboard
8 GLM-5 1 Feb 2026 Official leaderboard
#ModelScoreSettingDateSource
1 Claude Sonnet 4.6 2 Feb 2026 Official leaderboard
2 Claude Opus 4.6 2 Feb 2026 Official leaderboard
3 GPT-5.4 2 Mar 2026 Official leaderboard
4 Qwen3.5 397B-A17B 1 Feb 2026 Official leaderboard
5 Gemini 3.1 Pro 1 Feb 2026 Official leaderboard
6 DeepSeek-V3.2 1 Dec 2025 Official leaderboard
7 GLM-5 1 Feb 2026 Official leaderboard

Rankings by part

We read 13 parts from the benchmark’s own leaderboard, each ranking models on one part of the test, such as Story Quality, Overall User Experience and Relevance. Claude Sonnet 4.6 leads 5 of them, and 2 other models lead the rest. Arrows show how far a model moved from its overall rank among the models we track.

Story Quality

7 models we track have a score here.

  1. 1 Claude Sonnet 4.6 2
  2. 2 GPT-5.4 ↑1 2
  3. 3 Claude Opus 4.6 ↓1 2
  4. 4 Qwen3.5 397B-A17B 1
  5. 5 Gemini 3.1 Pro 1
  6. 6 GLM-5 ↑1 1
  7. 7 DeepSeek-V3.2 ↓1 1

Number one over time

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

Anthropic
Own leaderboard Not tested since May 2026

Source: Official NARRA-Gym leaderboard

Scores over time

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

Plot with other benchmarks
On the benchmark’s own leaderboard
1 1 2 2 2026 1 1 2 2 2026

Source: Official NARRA-Gym leaderboard

No lab model card reports it yet

We read the headline results of 540 launch posts from 57 labs, and none of them list this benchmark.

More in Writing and taste

All 65 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

Data audit

8 of 9 checks pass, 1 warning. Look first at the Newest score check. The newest came 8 months ago, in March 2026, for GPT-5.4. Every benchmark

Catalog record 2 pass
  • Pass Task access All tasks are public.
  • Pass Confidence Checked against a primary source.
Scores 1 warning 2 pass 2 notes
  • Pass Independent runs Scores from the official leaderboard (8 entries).
  • Pass Models scored 7 models we track, plus 1 other entry on its leaderboard.
  • Warning Newest score The newest came 8 months ago, in March 2026, for GPT-5.4.
  • Note Frontier models None of the top 10 has a score. The strongest tested is GPT-5.4 (#28).
  • Note Headroom Scored as a rating, with no fixed ceiling.
Leaderboard import 3 pass
  • Pass Importer We read the board on narra-gym.github.io: 7 rows for models we track and 1 other entry at the last read.
  • Pass Last read November 20th, 2026.
  • Pass Read worked The last read worked.
Consistency 1 pass 1 note
  • Note Lab claims No model has both a lab number and an independent run.
  • Pass Known issues No published audit found problems.

How we check