Benchmark Almanac

Open Tabletop GM Narrative Quality Probe v2

Qwen3-Next-80B-A3B-Instruct 4.88 leads since April 2026
Not in model cards
Version 2

Models narrate twelve tabletop scenes, including player failure, deception, and changes in tone. Five judges score each scene for atmosphere, character portrayal, and game management.

The maker’s table scores 37 models. Each model generated one response per scene. The five judges scored the same response.

Made by
Bobby Gray
Released
April 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
Interaction
Single turn
Run by
Anyone
Strongest model tested
Qwen3.5 397B-A17B #78 on the capability index
Last model added
April 2026
Scores refreshed
November 20th, 2026
37
Entries on its own leaderboard, 17 models we track

Qwen3-Next-80B-A3B-Instruct leads with 4.88 among models released by April 2026 Qwen3-Next-80B-A3B-Instruct leads with 4.88 among models released by April 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 April 2026, and nobody has tested a newer one. Among the 20 highest-ranked models on the capability index, these have no score yet:

  1. 1 Qwen3-Next-80B-A3B-Instruct 4.88
  2. 2 Mistral Medium 3.1 4.80
  3. 3 mistralai/ministral-8b-2512 4.76
  4. 4 Qwen3-235B-A22B 4.76
  5. 5 Gemma 3 27B 4.75
  6. 6 mistralai/mistral-large-2512 4.69
  7. 7 Nemotron 3 Nano 30B-A3B 4.68
  8. 8 Gemma 4 26B A4B 4.66
  9. 9 mistralai/mistral-small-3.2-24b-instruct 4.61
  10. 10 Qwen3.5 397B-A17B 4.59
  11. 11 Qwen3.5-122B-A10B 4.59
  12. 12 Qwen3.5-27B 4.56
  13. 13 Qwen3-32B 4.53
  14. 14 Gemma 4 31B IT 4.52
  15. 15 mistralai/mixtral-8x22b-instruct 4.51
  16. 16 thedrummer/cydonia-24b-v4.1 4.48
  17. 17 DeepSeek-V3.2 4.47
  18. 18 thedrummer/skyfall-36b-v2 4.45
  19. 19 Llama 4 Scout 4.45
  20. 20 mancer/weaver 4.43
  21. 21 nvidia/nemotron-3-super-120b-a12b 4.42
  22. 22 Llama 4 Maverick 4.41
  23. 23 Llama 3.3 70B Instruct 4.36
  24. 24 thedrummer/unslopnemo-12b 4.33
  25. 25 thedrummer/rocinante-12b 4.32
  26. 26 aion-labs/aion-rp-llama-3.1-8b 4.31
  27. 27 Hermes 4 405B 4.31
  28. 28 nousresearch/hermes-4-70b 4.25
  29. 29 sao10k/l3.1-70b-hanami-x1 4.22
  30. 30 sao10k/l3-lunaris-8b 4.18
  31. 31 sao10k/l3.1-euryale-70b 4.14
  32. 32 Qwen2.5-72B 4.10
  33. 33 anthracite-org/magnum-v4-72b 3.98
  34. 34 nousresearch/hermes-3-llama-3.1-405b 3.97
  35. 35 undi95/remm-slerp-l2-13b 3.82
  36. 36 gryphe/mythomax-l2-13b 3.67
  37. 37 sao10k/l3.3-euryale-70b 3.56

Source: Official Open Tabletop GM Narrative Quality Probe v2 leaderboard

#EntryScoreSettingDateSource
1 Qwen3-Next-80B-A3B-Instruct 4.88 Sep 2025 Official leaderboard
2 Mistral Medium 3.1 4.80 Aug 2025 Official leaderboard
3 mistralai/ministral-8b-2512 4.76 Official leaderboard
4 Qwen3-235B-A22B 4.76 Apr 2025 Official leaderboard
5 Gemma 3 27B 4.75 Mar 2025 Official leaderboard
6 mistralai/mistral-large-2512 4.69 Official leaderboard
7 Nemotron 3 Nano 30B-A3B 4.68 Dec 2025 Official leaderboard
8 Gemma 4 26B A4B 4.66 Apr 2026 Official leaderboard
9 mistralai/mistral-small-3.2-24b-instruct 4.61 Official leaderboard
10 Qwen3.5 397B-A17B 4.59 Feb 2026 Official leaderboard
11 Qwen3.5-122B-A10B 4.59 Feb 2026 Official leaderboard
12 Qwen3.5-27B 4.56 Feb 2026 Official leaderboard
13 Qwen3-32B 4.53 Apr 2025 Official leaderboard
14 Gemma 4 31B IT 4.52 Apr 2026 Official leaderboard
15 mistralai/mixtral-8x22b-instruct 4.51 Official leaderboard
16 thedrummer/cydonia-24b-v4.1 4.48 Official leaderboard
17 DeepSeek-V3.2 4.47 Dec 2025 Official leaderboard
18 thedrummer/skyfall-36b-v2 4.45 Official leaderboard
19 Llama 4 Scout 4.45 Apr 2025 Official leaderboard
20 mancer/weaver 4.43 Official leaderboard
21 nvidia/nemotron-3-super-120b-a12b 4.42 Official leaderboard
22 Llama 4 Maverick 4.41 Apr 2025 Official leaderboard
23 Llama 3.3 70B Instruct 4.36 Dec 2024 Official leaderboard
24 thedrummer/unslopnemo-12b 4.33 Official leaderboard
25 thedrummer/rocinante-12b 4.32 Official leaderboard
26 aion-labs/aion-rp-llama-3.1-8b 4.31 Official leaderboard
27 Hermes 4 405B 4.31 Aug 2025 Official leaderboard
28 nousresearch/hermes-4-70b 4.25 Official leaderboard
29 sao10k/l3.1-70b-hanami-x1 4.22 Official leaderboard
30 sao10k/l3-lunaris-8b 4.18 Official leaderboard
31 sao10k/l3.1-euryale-70b 4.14 Official leaderboard
32 Qwen2.5-72B 4.10 Sep 2024 Official leaderboard
33 anthracite-org/magnum-v4-72b 3.98 Official leaderboard
34 nousresearch/hermes-3-llama-3.1-405b 3.97 Official leaderboard
35 undi95/remm-slerp-l2-13b 3.82 Official leaderboard
36 gryphe/mythomax-l2-13b 3.67 Official leaderboard
37 sao10k/l3.3-euryale-70b 3.56 Official leaderboard
#ModelScoreSettingDateSource
1 Qwen3-Next-80B-A3B-Instruct 4.88 Sep 2025 Official leaderboard
2 Mistral Medium 3.1 4.80 Aug 2025 Official leaderboard
3 Qwen3-235B-A22B 4.76 Apr 2025 Official leaderboard
4 Gemma 3 27B 4.75 Mar 2025 Official leaderboard
5 Nemotron 3 Nano 30B-A3B 4.68 Dec 2025 Official leaderboard
6 Gemma 4 26B A4B 4.66 Apr 2026 Official leaderboard
7 Qwen3.5 397B-A17B 4.59 Feb 2026 Official leaderboard
8 Qwen3.5-122B-A10B 4.59 Feb 2026 Official leaderboard
9 Qwen3.5-27B 4.56 Feb 2026 Official leaderboard
10 Qwen3-32B 4.53 Apr 2025 Official leaderboard
11 Gemma 4 31B IT 4.52 Apr 2026 Official leaderboard
12 DeepSeek-V3.2 4.47 Dec 2025 Official leaderboard
13 Llama 4 Scout 4.45 Apr 2025 Official leaderboard
14 Llama 4 Maverick 4.41 Apr 2025 Official leaderboard
15 Llama 3.3 70B Instruct 4.36 Dec 2024 Official leaderboard
16 Hermes 4 405B 4.31 Aug 2025 Official leaderboard
17 Qwen2.5-72B 4.10 Sep 2024 Official leaderboard

Rankings by part

We read 3 parts from the benchmark’s own leaderboard: Atm, NPC and GM. Each ranks models on one part of the test. Qwen3-Next-80B-A3B-Instruct leads 2 of them, and Mistral Medium 3.1 leads the other one. Arrows show how far a model moved from its overall rank among the models we track.

Atm

17 models we track have a score here.

  1. 1 Qwen3-Next-80B-A3B-Instruct 4.95
  2. 2 Nemotron 3 Nano 30B-A3B ↑3 4.86
  3. 3 Qwen3-235B-A22B 4.84
  4. 4 Gemma 4 26B A4B ↑2 4.82
  5. 5 Gemma 3 27B ↓1 4.81
  6. 6 Mistral Medium 3.1 ↓4 4.78
  7. 7 Qwen3-32B ↑3 4.77
  8. 8 Qwen3.5 397B-A17B ↓1 4.75
  9. 9 Qwen3.5-27B 4.75
  10. 10 Qwen3.5-122B-A10B ↓2 4.71

Number one over time

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

Qwen
Own leaderboard Not tested since Apr 2026

Source: Official Open Tabletop GM Narrative Quality Probe v2 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
4.00 4.20 4.40 4.60 4.80 5.00 2025 2026 4.00 4.20 4.40 4.60 4.80 5.00 2025 2026

Source: Official Open Tabletop GM Narrative Quality Probe v2 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.

Versions

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 7 months ago, in April 2026, for Gemma 4 26B A4B. 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 (37 entries).
  • Pass Models scored 17 models we track, plus 20 other entries on its leaderboard.
  • Warning Newest score The newest came 7 months ago, in April 2026, for Gemma 4 26B A4B.
  • Note Frontier models None of the top 10 has a score. The strongest tested is Qwen3.5 397B-A17B (#78).
  • Note Headroom Scored as a plain number, with no fixed ceiling.
Leaderboard import 3 pass
  • Pass Importer We read the board on github.com: 17 rows for models we track and 20 other entries 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