Benchmark Almanac

FlavourBench Powered v43

Grok 4.6 73.4 leads since August 2026
Not in model cards
Version powered v43

Models choose ingredients for 640 tasks across four categories, including regional composition. Failed responses earn zero. Models must complete 608 tasks to qualify for ranking.

The archived table ranks 25 of 26 models. Claude Fable 5 completed 242 tasks and received no rank.

Released
August 2026
Kind
Taste and style
Level
Everyday
Scored
Checked automatically
Answers
Multiple choice
Tasks
All public
Measures
Capability
Language
English
In and out
Text → Text
Interaction
Single turn
Items from
Generated
Run by
Anyone
Strongest model tested
Claude Opus 5 #11 on the capability index
Last model added
August 2026
Scores refreshed
November 20th, 2026
25
Entries on its own leaderboard, 17 models we track

Grok 4.6 leads with 73.4 among models released by August 2026 Grok 4.6 leads with 73.4 among models released by August 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 August 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 Grok 4.6 73.4
  2. 2 Google: Gemini 3.1 Pro Preview 72.9
  3. 3 Muse Spark 1.2 72.4
  4. 4 GPT-5.6 Sol Pro 72.0
  5. 5 OpenAI: GPT-5.6 Terra Pro 71.5
  6. 6 Qwen 3.8 Max 71.4
  7. 7 Qwen3.8-2.4T-A95B 71.2
  8. 8 Claude Opus 5 70.8
  9. 9 Kimi K3 70.3
  10. 10 Gemini 3.6 Flash 70.1
  11. 11 Claude Sonnet 5 69.4
  12. 12 ByteDance Seed: Seed 2.1 Turbo 69.2
  13. 13 OpenAI: GPT-5.6 Luna Pro 69.2
  14. 14 Tencent Hy3 68.9
  15. 15 DeepSeek V4 Pro 0813 68.5
  16. 16 Meta: Muse Glimmer 30B 68.4
  17. 17 GLM-5.2 66.9
  18. 18 MiniMax-M3 66.7
  19. 19 NVIDIA: Nemotron 3.5 Lightning 65.2
  20. 20 Inkling 64.5
  21. 21 DeepSeek V4 Flash 0731 62.7
  22. 22 Mistral: Mistral Large 3 2512 61.4
  23. 23 Llama 4 Maverick 60.2
  24. 24 Command A+ 58.1
  25. 25 Cohere: Command R+ (08-2024) 36.7

Source: Official FlavourBench Powered v43 leaderboard

#EntryScore ±SettingDateSource
1 Grok 4.6 73.4 ±2.1 Aug 2026 Official leaderboard
2 Google: Gemini 3.1 Pro Preview 72.9 ±2.1 Official leaderboard
3 Muse Spark 1.2 72.4 ±2.2 Aug 2026 Official leaderboard
4 GPT-5.6 Sol Pro 72.0 ±2.2 Jul 2026 Official leaderboard
5 OpenAI: GPT-5.6 Terra Pro 71.5 ±2.2 Official leaderboard
6 Qwen 3.8 Max 71.4 ±2.2 Aug 2026 Official leaderboard
7 Qwen3.8-2.4T-A95B 71.2 ±2.2 Aug 2026 Official leaderboard
8 Claude Opus 5 70.8 ±2.2 Jul 2026 Official leaderboard
9 Kimi K3 70.3 ±2.2 Jul 2026 Official leaderboard
10 Gemini 3.6 Flash 70.1 ±2.2 Jul 2026 Official leaderboard
11 Claude Sonnet 5 69.4 ±2.2 Jun 2026 Official leaderboard
12 ByteDance Seed: Seed 2.1 Turbo 69.2 ±2.2 Official leaderboard
13 OpenAI: GPT-5.6 Luna Pro 69.2 ±2.2 Official leaderboard
14 Tencent Hy3 68.9 ±2.2 Jul 2026 Official leaderboard
15 DeepSeek V4 Pro 0813 68.5 ±2.4 Aug 2026 Official leaderboard
16 Meta: Muse Glimmer 30B 68.4 ±2.2 Official leaderboard
17 GLM-5.2 66.9 ±2.2 Jun 2026 Official leaderboard
18 MiniMax-M3 66.7 ±2.2 Jun 2026 Official leaderboard
19 NVIDIA: Nemotron 3.5 Lightning 65.2 ±2.3 Official leaderboard
20 Inkling 64.5 ±2.4 Jul 2026 Official leaderboard
21 DeepSeek V4 Flash 0731 62.7 ±2.3 Jul 2026 Official leaderboard
22 Mistral: Mistral Large 3 2512 61.4 ±2.3 Official leaderboard
23 Llama 4 Maverick 60.2 ±2.3 Apr 2025 Official leaderboard
24 Command A+ 58.1 ±2.2 May 2026 Official leaderboard
25 Cohere: Command R+ (08-2024) 36.7 ±2 Official leaderboard
#ModelScore ±SettingDateSource
1 Grok 4.6 73.4 ±2.1 Aug 2026 Official leaderboard
2 Muse Spark 1.2 72.4 ±2.2 Aug 2026 Official leaderboard
3 GPT-5.6 Sol Pro 72.0 ±2.2 Jul 2026 Official leaderboard
4 Qwen 3.8 Max 71.4 ±2.2 Aug 2026 Official leaderboard
5 Qwen3.8-2.4T-A95B 71.2 ±2.2 Aug 2026 Official leaderboard
6 Claude Opus 5 70.8 ±2.2 Jul 2026 Official leaderboard
7 Kimi K3 70.3 ±2.2 Jul 2026 Official leaderboard
8 Gemini 3.6 Flash 70.1 ±2.2 Jul 2026 Official leaderboard
9 Claude Sonnet 5 69.4 ±2.2 Jun 2026 Official leaderboard
10 Tencent Hy3 68.9 ±2.2 Jul 2026 Official leaderboard
11 DeepSeek V4 Pro 0813 68.5 ±2.4 Aug 2026 Official leaderboard
12 GLM-5.2 66.9 ±2.2 Jun 2026 Official leaderboard
13 MiniMax-M3 66.7 ±2.2 Jun 2026 Official leaderboard
14 Inkling 64.5 ±2.4 Jul 2026 Official leaderboard
15 DeepSeek V4 Flash 0731 62.7 ±2.3 Jul 2026 Official leaderboard
16 Llama 4 Maverick 60.2 ±2.3 Apr 2025 Official leaderboard
17 Command A+ 58.1 ±2.2 May 2026 Official leaderboard

Rankings by part

We read 4 parts from the benchmark’s own leaderboard, each ranking models on one part of the test, such as Constraint, Cultural composition and Pairing. Grok 4.6 leads 2 of them, and Claude Opus 5 leads the other 2. Arrows show how far a model moved from its overall rank among the models we track.

Constraint

17 models we track have a score here.

  1. 1 Grok 4.6 56.9
  2. 2 Muse Spark 1.2 55.6
  3. 3 Qwen 3.8 Max ↑1 53.4
  4. 4 Qwen3.8-2.4T-A95B ↑1 50.9
  5. 5 GPT-5.6 Sol Pro ↓2 50.7
  6. 6 Gemini 3.6 Flash ↑2 50.2
  7. 7 Kimi K3 48.3
  8. 8 DeepSeek V4 Pro 0813 ↑3 45.9
  9. 9 Tencent Hy3 ↑1 45.5
  10. 10 Claude Opus 5 ↓4 45.3

3 models have held number one on the official leaderboard since its release in August 2026

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

OpenAI Meta xAI
Own leaderboard Not tested since Aug 2026
Grok 4.6 73.4

Source: Official FlavourBench Powered v43 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
55.0 60.0 65.0 70.0 75.0 2025 2026 55.0 60.0 65.0 70.0 75.0 2025 2026

Source: Official FlavourBench Powered v43 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

All 9 checks pass. Every benchmark

Catalog record 2 pass
  • Pass Task access All tasks are public.
  • Pass Confidence Checked against a primary source.
Scores 3 pass 2 notes
  • Pass Independent runs Scores from the official leaderboard (25 entries).
  • Pass Models scored 17 models we track, plus 8 other entries on its leaderboard.
  • Pass Newest score The newest is from August 2026, for DeepSeek V4 Pro 0813.
  • Note Frontier models None of the top 10 has a score, which is fine for a retired test. The strongest tested is Claude Opus 5 (#11).
  • Note Headroom Scored as a plain number, with no fixed ceiling.
Leaderboard import 3 pass
  • Pass Importer We read the board on huggingface.co: 17 rows for models we track and 8 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