Skip to main content
BenchLM
Data

LisanBench

We show this table for reference; we do not rank on it.

A word-chain reasoning benchmark that tests planning, recall, constraint following, and vocabulary depth by asking models to extend non-repeating edit-distance-1 chains.

Difficulty-weighted score on LisanBench — November 12, 2026 snapshot

We mirror the published difficulty-weighted score view for LisanBench. Muse Wonder Full (max) leads the public snapshot at 11089.05, followed by Muse Wonder Full (xhigh) (9337.23) and Muse Wonder (max) (8303.13). We do not use these results to rank models overall.

1

Muse Wonder Full (max)

Meta

11089.05
2

Muse Wonder Full (xhigh)

Meta

9337.23
3

Muse Wonder (max)

Meta

8303.13

211 modelsReasoningCurrentDisplay onlyUpdated November 12, 2026 snapshot

Difficulty-weighted score table (211 models)

Difficulty-weighted score results for LisanBench
RankModel / configurationScoreParameters (B)Open / closed
1Muse Wonder Full (max)Meta
11089.05
Not reportedUnknown
2Muse Wonder Full (xhigh)Meta
9337.23
Not reportedUnknown
3Muse Wonder (max)Meta
8303.13
Not reportedUnknown
4DeepSeek V4.1 Flash 1017DeepSeek
7732.61
Not reportedOpen
5Muse Wonder Full (high)Meta
6934.17
Not reportedUnknown
6Muse Wonder (xhigh)Meta
6601.81
Not reportedUnknown
7Claude Opus 5.5Anthropic
5963.10
Not reportedClosed
8Muse Wonder (high)Meta
5377.79
Not reportedUnknown
9DeepSeek V4.1 Flash 1017DeepSeek
5332.68
Not reportedOpen
10Muse Wonder Full (medium)Meta
5224.05
Not reportedUnknown
11Claude Opus 4.7 (Adaptive)Anthropic
5122.60
Not reportedClosed
12Claude Opus 5Anthropic
5068.25
Not reportedClosed
13Claude Fable 5.1Anthropic
4573.41
Not reportedClosed
14Claude Fable 5Anthropic
4561.82
Not reportedClosed
15DeepSeek V4.1 Flash 1017DeepSeek
4088.31
Not reportedOpen
16GPT-6.1 SolOpenAI
4004.30
Not reportedClosed
17Muse Wonder (medium)Meta
3909.43
Not reportedUnknown
18Gemini 4 ArgonGoogle
3849.26
Not reportedClosed
19Muse Wonder Full (low)Meta
3822.58
Not reportedUnknown
20Claude Opus 4.6 (Adaptive)Anthropic
3526.49
Not reportedClosed
21Kimi K3Moonshot AI
3525.77
Not reportedOpen
22GPT-5.5OpenAI
3315.52
Not reportedClosed
23GPT-6 SolOpenAI
3118.02
Not reportedClosed
24GPT-6 AstraOpenAI
3112.84
Not reportedClosed
25Claude Sonnet 5.5Anthropic
2983.19
Not reportedClosed
26Claude Sonnet 4.6Anthropic
2944.27
Not reportedClosed
27GPT-5.4OpenAI
2738.16
Not reportedClosed
28Claude Haiku 5.5Anthropic
2711.25
Not reportedClosed
29Claude Opus 4.8Anthropic
2693.67
Not reportedClosed
30Muse Wonder (low)Meta
2679.33
Not reportedUnknown
31Claude Opus 5Anthropic
2270.40
Not reportedClosed
32Claude Opus 4.5 ThinkingAnthropic
2204.43
Not reportedClosed
33GPT-6.1 SolOpenAI
2187.04
Not reportedClosed
34GPT-6 SolOpenAI
2126.57
Not reportedClosed
35Gemini 3.1 ProGoogle
1929.11
Not reportedClosed
36Claude Opus 5.5Anthropic
1926.73
Not reportedClosed
37Claude Fable 5.1Anthropic
1810.97
Not reportedClosed
38Grok 4xAI
1778.40
Not reportedClosed
39Claude Sonnet 5Anthropic
1736.56
Not reportedClosed
40GPT-6 AstraOpenAI
1666.01
Not reportedClosed
41Muse Wonder Full (minimal)Meta
1563.86
Not reportedUnknown
42Claude Opus 5.5Anthropic
1532.67
Not reportedClosed
43o3OpenAI
1523.00
Not reportedClosed
44Claude Haiku 5.5Anthropic
1517.65
Not reportedClosed
45Deepseek V3.2 Speciale (thinking)DeepSeek
1510.70
Not reportedUnknown
46Claude Fable 5.1Anthropic
1482.22
Not reportedClosed
47Grok 4.20xAI
1464.63
Not reportedClosed
48GPT-5.2OpenAI
1458.80
Not reportedClosed
49GPT-5 (medium)OpenAI
1457.28
Not reportedClosed
50GPT-5.6 SolOpenAI
1403.28
Not reportedClosed
51Muse Wonder (minimal)Meta
1395.33
Not reportedUnknown
52GPT-5.6 LunaOpenAI
1356.30
Not reportedClosed
53GPT-6 LunaOpenAI
1319.88
Not reportedClosed
54Claude Sonnet 5Anthropic
1268.24
Not reportedClosed
55GPT-6.1 SolOpenAI
1245.43
Not reportedClosed
56Claude Sonnet 5.5Anthropic
1229.79
Not reportedClosed
57GPT-5.6 TerraOpenAI
1196.90
Not reportedClosed
58Gemini 3 ProGoogle
1130.63
Not reportedClosed
59Gemini 3.5 FlashGoogle
1128.26
Not reportedClosed
60GPT-6 SolOpenAI
1121.00
Not reportedClosed
61GPT-5.6 LunaOpenAI
1112.53
Not reportedClosed
62Claude Sonnet 4.5 ThinkingAnthropic
1090.82
Not reportedClosed
63DeepSeek V4 Flash 0731DeepSeek
1063.47
Not reportedOpen
64DeepSeek V4 Pro 0813DeepSeek
1059.51
Not reportedOpen
65Claude Haiku 5.5Anthropic
980.94
Not reportedClosed
66GPT-6 Luna (medium)OpenAI
958.42
Not reportedUnknown
67DeepSeek V3.2 (Thinking)DeepSeek
925.31
Not reportedOpen
68Gemini 3.1 ProGoogle
872.67
Not reportedClosed
69Step 3.5 FlashStepFun
811.21
Not reportedOpen
70Grok 4 Fast (Reasoning)xAI
806.48
Not reportedClosed
71GPT-5 miniOpenAI
758.84
Not reportedClosed
72Claude Opus 5.5Anthropic
729.77
Not reportedClosed
73GPT-5.4 miniOpenAI
656.62
Not reportedClosed
74GPT-5.6 LunaOpenAI
648.18
Not reportedClosed
75Kimi K2.5 (Reasoning)Moonshot AI
641.96
Not reportedOpen
76GPT-5.4 miniOpenAI
634.44
Not reportedClosed
77GPT-5 nanoOpenAI
608.60
Not reportedClosed
78Grok 4.1 Fast (Reasoning)xAI
604.43
Not reportedClosed
79Sonnet 4 (16k)Anthropic
602.95
Not reportedUnknown
80GLM-5.2Z.AI
591.99
Not reportedOpen
81Gemini 3 FlashGoogle
591.88
Not reportedClosed
82GPT-5.4 miniOpenAI
591.48
Not reportedClosed
83Gemini 4 ArgonGoogle
590.93
Not reportedClosed
84GPT-5.4 nanoOpenAI
543.21
Not reportedClosed
85Claude Fable 5.1Anthropic
529.07
Not reportedClosed
86o3-miniOpenAI
518.37
Not reportedClosed
87Gemini 4 ArgonGoogle
479.28
Not reportedClosed
88GPT-6 Luna (low)OpenAI
470.31
Not reportedUnknown
89Doubao Seed 2.0 Pro (thinking)StepFun
456.93
Not reportedUnknown
90GPT-5.6 Luna (low)OpenAI
448.40
Not reportedClosed
91GPT-OSS 120BOpenAI
448.33
Not reportedOpen
92Qwen3.5 397B (Reasoning)Alibaba
387.74
Not reportedOpen
93Claude Sonnet 5.5Anthropic
382.78
Not reportedClosed
94Claude Sonnet 5.5Anthropic
379.86
Not reportedClosed
95Claude Opus 5Anthropic
366.79
Not reportedClosed
96o4-miniOpenAI
352.68
Not reportedClosed
97GLM-5 (Reasoning)Z.AI
336.34
Not reportedOpen
98GPT-5.6 SolOpenAI
332.46
Not reportedClosed
99GPT-5.4 miniOpenAI
331.79
Not reportedClosed
100Claude Sonnet 5.5Anthropic
312.58
Not reportedClosed
101GPT-5.5OpenAI
305.30
Not reportedClosed
102GPT-6 SolOpenAI
290.00
Not reportedClosed
103GPT-5.4 miniOpenAI
282.67
Not reportedClosed
104Claude Opus 4.8Anthropic
270.10
Not reportedClosed
105Doubao Seed 2.0 Lite (thinking)StepFun
265.12
Not reportedUnknown
106Opus 4Anthropic
262.56
Not reportedUnknown
107Doubao Seed 1.8 (thinking)StepFun
255.84
Not reportedUnknown
108MiniMax M2.5MiniMax
228.38
Not reportedClosed
109Qwen3 235B 2507 (Reasoning)Alibaba
226.22
Not reportedOpen
110Claude Opus 4.7Anthropic
217.52
Not reportedClosed
111Claude 4.1 OpusAnthropic
215.66
Not reportedClosed
112Claude Sonnet 4.6Anthropic
208.64
Not reportedClosed
113Claude Sonnet 5Anthropic
208.21
Not reportedClosed
114Claude Haiku 5.5Anthropic
197.89
Not reportedClosed
115Gemini 2.5 ProGoogle
197.43
Not reportedClosed
116Grok 3 MinixAI
191.37
Not reportedClosed
117Grok 3 [Beta]xAI
188.12
Not reportedClosed
118Claude Haiku 5.5Anthropic
178.44
Not reportedClosed
119Claude 3.7 SonnetAnthropic
162.44
Not reportedClosed
120GPT-OSS 20BOpenAI
156.99
Not reportedOpen
121GPT-5.6 TerraOpenAI
156.14
Not reportedClosed
122Doubao Seed 2.0 Mini (thinking)StepFun
150.35
Not reportedUnknown
123Claude 4 SonnetAnthropic
150.17
Not reportedClosed
124Claude 3.5 SonnetAnthropic
149.65
Not reportedClosed
125Claude 3.5 SonnetAnthropic
129.73
Not reportedClosed
126GPT-6 LunaOpenAI
122.65
Not reportedClosed
127Gemini 1.5 ProGoogle
119.72
Not reportedClosed
128DeepSeek V3.2DeepSeek
119.04
Not reportedOpen
129DeepSeek V4 Pro 0813DeepSeek
117.14
Not reportedOpen
130Gemini 2.5 FlashGoogle
112.75
Not reportedClosed
131DeepSeek-R1-0528DeepSeek
111.60
Not reportedOpen
132Qwen3.5-122B-A10BAlibaba
109.62
Not reportedOpen
133GPT-5.4OpenAI
109.30
Not reportedClosed
134GLM-4.5Z.AI
108.32
Not reportedClosed
135Qwen3.5-35B-A3BAlibaba
107.61
Not reportedOpen
136Olmo 3 32B (thinking)Allen AI
104.86
Not reportedUnknown
137Claude Sonnet 4.5Anthropic
103.58
Not reportedClosed
138DeepSeek V3DeepSeek
103.39
Not reportedOpen
139o1-miniOpenAI
103.10
Not reportedClosed
140DeepSeek V4.1 Flash 1017DeepSeek
101.81
Not reportedOpen
141GPT-5.6 LunaOpenAI
96.63
Not reportedClosed
142GPT-4oOpenAI
94.16
Not reportedClosed
143Claude Opus 4.5Anthropic
93.49
Not reportedClosed
144GPT-4 TurboOpenAI
91.48
Not reportedClosed
145Kimi K2Moonshot AI
85.50
Not reportedClosed
146Claude Opus 4.6Anthropic
85.47
Not reportedClosed
147Qwen3 4B (16k)Alibaba
77.21
Not reportedUnknown
148Claude 3 OpusAnthropic
75.77
Not reportedClosed
149Gemini 2.5 FlashGoogle
72.17
Not reportedClosed
150MiniMax M1 80kMiniMax
66.71
Not reportedClosed
151GLM-5.2Z.AI
64.50
Not reportedOpen
152Gemini 2.0 FlashGoogle
61.61
Not reportedClosed
153DeepSeek V4 Flash 0731DeepSeek
56.78
Not reportedOpen
154Horizon BetaOpenRouter
55.69
Not reportedUnknown
155Gemini Flash 1.5Google
55.20
Not reportedUnknown
156GLM-4.5-AirZ.AI
54.86
Not reportedClosed
157Nova ProAmazon
54.38
Not reportedClosed
158GLM-4.7Z.AI
54.27
Not reportedOpen
159Polaris AlphaOpenRouter
53.34
Not reportedUnknown
160Claude Haiku 4.5Anthropic
52.64
Not reportedClosed
161GPT-3.5-turbo-0613 (RAGTruth generator)RAGTruth authors
50.96
Not reportedUnknown
162Qwen3 CoderAlibaba
50.95
Not reportedUnknown
163Llama 3.1 405BMeta
48.88
Not reportedOpen
164Grok 4.1 FastxAI
47.29
Not reportedClosed
165GLM-4.6Z.AI
44.20
Not reportedOpen
166Gemma 3 27BGoogle
43.79
Not reportedOpen
167Mistral Medium 3Mistral
43.31
Not reportedClosed
168GPT-5.4 miniOpenAI
42.74
Not reportedClosed
169Llama 4 MaverickMeta
42.33
Not reportedOpen
170GPT-4.1OpenAI
42.02
Not reportedClosed
171Ernie 4.5 300B A47BBaidu
40.99
Not reportedUnknown
172Sherlock Dash AlphaOpenRouter
40.07
Not reportedUnknown
173Devstral MediumMistral AI
40.03
Not reportedUnknown
174Gemini 2.0 Flash Lite 001Google
38.77
Not reportedUnknown
175Claude 3.5 HaikuAnthropic
38.22
Not reportedClosed
176Gemini 2.5 Flash-LiteGoogle
38.21
Not reportedClosed
177Llama 3.1 70BMeta
38.06
Not reportedOpen
178Qwen3 1.7B (16k)Alibaba
38.06
Not reportedUnknown
179Claude 3 HaikuAnthropic
35.46
Not reportedClosed
180Mistral Large 2411Mistral AI
34.64
Not reportedUnknown
181GPT-4.1 miniOpenAI
32.82
Not reportedClosed
182Gemini 2.5 Flash-LiteGoogle
32.37
Not reportedClosed
183Qwen3 235B A22B 2507Alibaba
31.92
Not reportedUnknown
184Llama 4 ScoutMeta
30.68
Not reportedOpen
185MiMo-V2-FlashXiaomi
28.41
Not reportedOpen
186Nova Lite V1Amazon
25.24
Not reportedUnknown
187Nova Micro V1Amazon
24.88
Not reportedUnknown
188Gemini Flash 1.5 8BGoogle
24.38
Not reportedUnknown
189Qwen3 32BAlibaba
23.86
Not reportedOpen
190Gemma 3 12BGoogle
21.56
Not reportedUnknown
191GPT-4o miniOpenAI
20.33
Not reportedClosed
192Qwen3 30B A3B 2507Alibaba
18.98
Not reportedUnknown
193Mistral Small 3.2 24BMistral AI
17.77
Not reportedUnknown
194Qwen3 14BAlibaba
16.68
Not reportedUnknown
195Qwen3 8BAlibaba
15.50
Not reportedUnknown
196GPT-4.1 nanoOpenAI
14.95
Not reportedClosed
197Devstral SmallMistral
13.22
Not reportedOpen
198Codestral 2508Mistral AI
13.15
Not reportedUnknown
199Ministral 14B 2512Mistral AI
12.29
Not reportedUnknown
200Ministral 8B 2512Mistral AI
11.78
Not reportedUnknown
201Mistral NemoMistral AI
11.47
Not reportedUnknown
202GPT-5.4 nanoOpenAI
11.16
Not reportedClosed
203Gemma 3 4BGoogle
9.41
Not reportedUnknown
204Qwen3 0.6B (16k)Alibaba
9.17
Not reportedUnknown
205Qwen3 4BAlibaba
7.90
Not reportedUnknown
206Ministral 3B 2512Mistral AI
6.53
Not reportedUnknown
207Qwen3 1.7BAlibaba
6.27
Not reportedUnknown
208Llama 3.1 8BMeta
3.92
Not reportedOpen
209Llama 3.2 3BMeta
2.85
Not reportedUnknown
210Llama 3.2 1BMeta
0.63
Not reportedUnknown
211Qwen3 0.6BAlibaba
0.06
Not reportedUnknown

How LisanBench is shown here

BenchLM mirrors the current public LisanBench difficulty-weighted leaderboard using the official dataset published at lisanbench.com for the November 12, 2026 snapshot. The public benchmark tests 211 model variants across 50 starting words, with 3 trials per starting word.

LisanBench is a strong reasoning reference, but BenchLM currently keeps it display only rather than weighted. The public leaderboard is highly variant-specific, strongly English-vocabulary-dependent, and not yet aligned cleanly enough with BenchLM canonical model rows to use as a ranking input.

Snapshot

211 model variants50 starting words3 trials per wordDifficulty-weighted scoresDisplay only
LisanBench leaderboardLisanBench methodologyCode and data

The published LisanBench snapshot places Muse Wonder Full (max) first at 11089.05. The third row is 2785.92 score units behind. The broader top-10 range is 5865.01 score units, so the table still separates the published systems.

211 models have been evaluated on LisanBench. The benchmark falls in the Reasoning category. LisanBench is currently displayed for reference but excluded from the scoring formula, so it does not directly affect overall rankings.

About LisanBench

Year

2026

Tasks

50 starting words × 3 trials

Format

Difficulty-weighted word-chain reasoning

Difficulty

Open-ended lexical planning

BenchLM mirrors the public difficulty-weighted LisanBench leaderboard as a display-only reasoning benchmark. The public benchmark currently evaluates 128 model variants across 50 starting words with 3 trials per word.

LisanBench methodologyPublic benchmark source

Freshness and provenance

Version

LisanBench 2026

Refresh cadence

Static

Staleness state

Current

Question availability

Public benchmark set

CurrentDisplay only

BenchLM uses freshness metadata to decide whether a benchmark should still be treated as a strong differentiator, a benchmark to watch, or a display-only reference. For the full scoring policy, see the BenchLM methodology page.

Questions

What does LisanBench measure?

A word-chain reasoning benchmark that tests planning, recall, constraint following, and vocabulary depth by asking models to extend non-repeating edit-distance-1 chains.

Which model leads the published LisanBench snapshot?

Muse Wonder Full (max) currently leads the published LisanBench snapshot with 11089.05 difficulty-weighted score. BenchLM shows this benchmark for display only and does not use it in overall rankings.

How many models are evaluated on LisanBench?

The November 12, 2026 snapshot snapshot contains 211 AI models.

Last updated: November 12, 2026 snapshot · mirrored from the public benchmark leaderboard

Know when it’s worth switching models

The model to choose, the cheaper alternative, and the release we would wait on.

Read a sample issue

Join 7,000+ subscribers.

One email each week. Unsubscribe anytime.