Paraphrase Stability Benchmark
Models respond across a grid of paraphrases and contexts. The harness reports its Stability Agreement Average with a confidence interval.
Kaggle publishes scores for 25 distinct model versions.
- Made by
- Chauhan Balaji
- Released
- October 2026
- Kind
- Taste and style
- Level
- School
- Scored
- Checked automatically
- Answers
- Open-ended
- Tasks
- All public
- Measures
- Propensity
- Behavior
- Unreliability
- Language
- English
- Interaction
- Single turn
- Refresh
- One fixed set
- Run by
- Anyone
- Strongest model tested
- GPT-6 Astra #4 on the capability index
- Last model added
- September 2026
- Scores refreshed
- November 20th, 2026
4 models share the top score of 1.00 among models released by September 2026 4 models share the top score of 1.00 among models released by September 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 September 2026, and nobody has tested a newer one. Among the 20 highest-ranked models on the capability index, these have no score yet:
- 1.00 GLM-5
- 1.00 GPT-5.6 Sol
- 1.00 Claude Opus 5.5
- 1.00 GPT-6 Sol
- 0.97
Gemini 3.1 Flash-Lite
- 0.97
Gemma 4 26B A4B
- 0.97 Claude Opus 4.8
- 0.97 GPT-6 Astra
- 0.97 GPT-6 Luna
- 0.97
Gemini 3 Flash
- 0.96 GPT-6.1 Sol
- 0.94
Gemini 3.8 Flash
- 0.94
Gemini 3.6 Flash
- 0.93
Gemini 2.5 Pro
- 0.93 Claude Sonnet 5
- 0.91 Claude Opus 5
- 0.91 Qwen 3 Next 80B Instruct
- 0.90
Gemini 3.5 Flash
- 0.90
Gemini 3.1 Pro
- 0.88 GPT-5.4
- 0.87
Gemini 2.5 Flash
- 0.85 Claude Sonnet 5.5
- 0.84
Gemini 3.5 Flash-Lite
- 0.72 GPT-5.4 Mini
- 0.66 Grok 4.20
- 1 GLM-5 1.00
- 2 GPT-5.6 Sol 1.00
- 3 Claude Opus 5.5 1.00
- 4 GPT-6 Sol 1.00
- 5
Gemini 3.1 Flash-Lite 0.97
- 6
Gemma 4 26B A4B 0.97
- 7 Claude Opus 4.8 0.97
- 8 GPT-6 Astra 0.97
- 9 GPT-6 Luna 0.97
- 10
Gemini 3 Flash 0.97
- 11 GPT-6.1 Sol 0.96
- 12
Gemini 3.8 Flash 0.94
- 13
Gemini 3.6 Flash 0.94
- 14
Gemini 2.5 Pro Jun 2025 0.93
- 15 Claude Sonnet 5 0.93
- 16 Claude Opus 5 0.91
- 17 Qwen 3 Next 80B Instruct 0.91
- 18
Gemini 3.5 Flash 0.90
- 19
Gemini 3.1 Pro 0.90
- 20 GPT-5.4 0.88
- 21
Gemini 2.5 Flash Sep 2025 0.87
- 22 Claude Sonnet 5.5 0.85
- 23
Gemini 3.5 Flash-Lite 0.84
- 24 GPT-5.4 Mini 0.72
- 25 Grok 4.20 reasoning 0.66
Source: Official Paraphrase Stability Benchmark leaderboard
- 1.00 GLM-5
- 1.00 GPT-5.6 Sol
- 1.00 Claude Opus 5.5
- 1.00 GPT-6 Sol
- 0.97
Gemini 3.1 Flash-Lite
- 0.97
Gemma 4 26B A4B
- 0.97 Claude Opus 4.8
- 0.97 GPT-6 Astra
- 0.97 GPT-6 Luna
- 0.97
Gemini 3 Flash
- 0.96 GPT-6.1 Sol
- 0.94
Gemini 3.8 Flash
- 0.94
Gemini 3.6 Flash
- 0.93
Gemini 2.5 Pro
- 0.93 Claude Sonnet 5
- 0.91 Claude Opus 5
- 0.90
Gemini 3.5 Flash
- 0.90
Gemini 3.1 Pro
- 0.88 GPT-5.4
- 0.87
Gemini 2.5 Flash
- 0.85 Claude Sonnet 5.5
- 0.84
Gemini 3.5 Flash-Lite
- 0.72 GPT-5.4 Mini
- 0.66 Grok 4.20
- 1 GLM-5 1.00
- 2 GPT-5.6 Sol 1.00
- 3 Claude Opus 5.5 1.00
- 4 GPT-6 Sol 1.00
- 5
Gemini 3.1 Flash-Lite 0.97
- 6
Gemma 4 26B A4B 0.97
- 7 Claude Opus 4.8 0.97
- 8 GPT-6 Astra 0.97
- 9 GPT-6 Luna 0.97
- 10
Gemini 3 Flash 0.97
- 11 GPT-6.1 Sol 0.96
- 12
Gemini 3.8 Flash 0.94
- 13
Gemini 3.6 Flash 0.94
- 14
Gemini 2.5 Pro Jun 2025 0.93
- 15 Claude Sonnet 5 0.93
- 16 Claude Opus 5 0.91
- 17
Gemini 3.5 Flash 0.90
- 18
Gemini 3.1 Pro 0.90
- 19 GPT-5.4 0.88
- 20
Gemini 2.5 Flash Sep 2025 0.87
- 21 Claude Sonnet 5.5 0.85
- 22
Gemini 3.5 Flash-Lite 0.84
- 23 GPT-5.4 Mini 0.72
- 24 Grok 4.20 reasoning 0.66
Source: Official Paraphrase Stability Benchmark leaderboard
| # | Entry | Score | Setting | Released | Run on | Source |
|---|---|---|---|---|---|---|
| 1 | GLM-5 | 1.00 | Feb 2026 | Oct 2026 | Official leaderboard | |
| 2 | GPT-5.6 Sol | 1.00 | Jul 2026 | Oct 2026 | Official leaderboard | |
| 3 | Claude Opus 5.5 | 1.00 | Sep 2026 | Oct 2026 | Official leaderboard | |
| 4 | GPT-6 Sol | 1.00 | Sep 2026 | Oct 2026 | Official leaderboard | |
| 5 | | 0.97 | Mar 2026 | Oct 2026 | Official leaderboard | |
| 6 | | 0.97 | Apr 2026 | Oct 2026 | Official leaderboard | |
| 7 | Claude Opus 4.8 | 0.97 | May 2026 | Oct 2026 | Official leaderboard | |
| 8 | GPT-6 Astra | 0.97 | Sep 2026 | Oct 2026 | Official leaderboard | |
| 9 | GPT-6 Luna | 0.97 | Sep 2026 | Oct 2026 | Official leaderboard | |
| 10 | | 0.97 | Dec 2025 | Oct 2026 | Official leaderboard | |
| 11 | GPT-6.1 Sol | 0.96 | Sep 2026 | Oct 2026 | Official leaderboard | |
| 12 | | 0.94 | Sep 2026 | Oct 2026 | Official leaderboard | |
| 13 | | 0.94 | Jul 2026 | Oct 2026 | Official leaderboard | |
| 14 | | 0.93 | Jun 2025 | Oct 2026 | Official leaderboard | |
| 15 | Claude Sonnet 5 | 0.93 | Jun 2026 | Oct 2026 | Official leaderboard | |
| 16 | Claude Opus 5 | 0.91 | Jul 2026 | Oct 2026 | Official leaderboard | |
| 17 | Qwen 3 Next 80B Instruct | 0.91 | Oct 2026 | Official leaderboard | ||
| 18 | | 0.90 | May 2026 | Oct 2026 | Official leaderboard | |
| 19 | | 0.90 | Feb 2026 | Oct 2026 | Official leaderboard | |
| 20 | GPT-5.4 | 0.88 | Mar 2026 | Oct 2026 | Official leaderboard | |
| 21 | | 0.87 | Sep 2025 | Oct 2026 | Official leaderboard | |
| 22 | Claude Sonnet 5.5 | 0.85 | Sep 2026 | Oct 2026 | Official leaderboard | |
| 23 | | 0.84 | Jul 2026 | Oct 2026 | Official leaderboard | |
| 24 | GPT-5.4 Mini | 0.72 | Mar 2026 | Oct 2026 | Official leaderboard | |
| 25 | Grok 4.20 | 0.66 | reasoning | Feb 2026 | Oct 2026 | Official leaderboard |
| # | Model | Score | Setting | Released | Run on | Source |
|---|---|---|---|---|---|---|
| 1 | GLM-5 | 1.00 | Feb 2026 | Oct 2026 | Official leaderboard | |
| 2 | GPT-5.6 Sol | 1.00 | Jul 2026 | Oct 2026 | Official leaderboard | |
| 3 | Claude Opus 5.5 | 1.00 | Sep 2026 | Oct 2026 | Official leaderboard | |
| 4 | GPT-6 Sol | 1.00 | Sep 2026 | Oct 2026 | Official leaderboard | |
| 5 | | 0.97 | Mar 2026 | Oct 2026 | Official leaderboard | |
| 6 | | 0.97 | Apr 2026 | Oct 2026 | Official leaderboard | |
| 7 | Claude Opus 4.8 | 0.97 | May 2026 | Oct 2026 | Official leaderboard | |
| 8 | GPT-6 Astra | 0.97 | Sep 2026 | Oct 2026 | Official leaderboard | |
| 9 | GPT-6 Luna | 0.97 | Sep 2026 | Oct 2026 | Official leaderboard | |
| 10 | | 0.97 | Dec 2025 | Oct 2026 | Official leaderboard | |
| 11 | GPT-6.1 Sol | 0.96 | Sep 2026 | Oct 2026 | Official leaderboard | |
| 12 | | 0.94 | Sep 2026 | Oct 2026 | Official leaderboard | |
| 13 | | 0.94 | Jul 2026 | Oct 2026 | Official leaderboard | |
| 14 | | 0.93 | Jun 2025 | Oct 2026 | Official leaderboard | |
| 15 | Claude Sonnet 5 | 0.93 | Jun 2026 | Oct 2026 | Official leaderboard | |
| 16 | Claude Opus 5 | 0.91 | Jul 2026 | Oct 2026 | Official leaderboard | |
| 17 | | 0.90 | May 2026 | Oct 2026 | Official leaderboard | |
| 18 | | 0.90 | Feb 2026 | Oct 2026 | Official leaderboard | |
| 19 | GPT-5.4 | 0.88 | Mar 2026 | Oct 2026 | Official leaderboard | |
| 20 | | 0.87 | Sep 2025 | Oct 2026 | Official leaderboard | |
| 21 | Claude Sonnet 5.5 | 0.85 | Sep 2026 | Oct 2026 | Official leaderboard | |
| 22 | | 0.84 | Jul 2026 | Oct 2026 | Official leaderboard | |
| 23 | GPT-5.4 Mini | 0.72 | Mar 2026 | Oct 2026 | Official leaderboard | |
| 24 | Grok 4.20 | 0.66 | reasoning | Feb 2026 | Oct 2026 | Official leaderboard |
Number one over time
Lab claims count from the day the card came out, independent runs from the model’s release.
Source: Official Paraphrase Stability Benchmark leaderboard
Scores over time
All the scores we have, one dot each, so a model tested at several settings appears more than once.
Source: Official Paraphrase Stability Benchmark 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
- 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 10 checks pass. Every benchmark
Catalog record 2 pass
- Pass Task access All tasks are public.
- Pass Confidence Checked against a primary source.
Scores 4 pass 1 note
- Pass Independent runs Scores from the official leaderboard (25 entries).
- Pass Models scored 24 models we track, plus 1 other entry on its leaderboard.
- Pass Newest score The newest is from September 2026, for GPT-6.1 Sol.
- Pass Frontier models 4 of the top 10 have a score, led by GPT-6 Astra (#4).
- Note Headroom Scored as a plain number, with no fixed ceiling.
Leaderboard import 3 pass
- Pass Importer We read the board on kaggle.com: 24 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.