Skip to main content

Benchmark profile

Testing the Limits of Chain-of-thought with Multistep Soft Reasoning (MuSR)

A dataset for evaluating language models on multistep soft reasoning tasks specified in natural language narratives. Tests the ability to perform complex, structured reasoning.

Data verified

How BenchLM shows MuSR right now

BenchLM is tracking MuSR in the local dataset, but exact-source verification records for these rows are still being attached. To avoid a blank benchmark page, BenchLM shows the current tracked rows below as a display-only reference table.

These tracked rows are useful for inspection and spot-checking, but until exact-source attachments are completed they should not be treated as fully verified public benchmark rows.

113 tracked modelsLocal tracked rowsAwaiting exact-source attachmentsDisplay only

Tracked score on MuSR — July 20, 2026

BenchLM mirrors the published tracked score view for MuSR. GPT-5.2 Pro leads the public snapshot at 95% , followed by GPT-5.4 (94%) and GPT-5.3 Instant (94%). BenchLM does not use these results to rank models overall.

113 modelsReasoningStaleDisplay onlyUpdated July 20, 2026

Tracked score table (113 models)

Score
1
95%
2
GPT-5.4OpenAI
94%
3
94%
4
93%
5
93%
6
93%
8
93%
9
93%
10
GPT-5.2OpenAI
93%
11
93%
12
93%
13
93%
14
93%
17
GPT-5.1OpenAI
91%
19
89%
21
87%
22
86%
23
86%
24
85%
26
o3-proOpenAI
84%
27
82%
28
o3OpenAI
82%
29
GLM-5Z.AI
82%
30
82%
31
82%
32
Mercury 2Inception
82%
33
81%
35
80%
36
Qwen2.5-1MAlibaba
79%
37
79%
38
79%
39
78%
40
78%
41
78%
42
76%
43
75%
44
75%
45
74%
46
Aion-2.0Aion Labs
74%
47
Kimi K2.5Moonshot AI
72%
48
72%
50
71%
52
69%
53
Seed 1.6ByteDance
69%
54
68%
56
Seed-2.0-LiteByteDance
66%
57
65%
58
64%
59
64%
60
63%
61
GPT-4oOpenAI
62%
63
61%
64
61%
65
61%
67
60%
69
59%
71
58%
72
Claude 3 OpusAnthropic
57%
73
Seed-2.0-MiniByteDance
57%
74
56.2%
75
56%
76
54%
77
52%
79
50.8%
80
50%
81
50%
82
Moonshot v1Moonshot AI
49%
83
48%
84
47%
85
46%
86
45.1%
87
45.1%
90
43%
92
LFM2-24B-A2BLiquidAI
42%
93
41.4%
94
41%
95
DeepSeek-R1DeepSeek
40%
97
Nova ProAmazon
37%
99
35%
100
33%
102
32%
103
31%
104
31%
106
29%
107
27%
108
26%
109
26%
111
25%
112
22%
113
20%

The published MuSR snapshot places GPT-5.2 Pro first at 95%. The third row is 1.0 points behind. The broader top-10 range is 2.0 points, so many of the published results sit in a relatively narrow band.

113 models have been evaluated on MuSR. The benchmark falls in the Reasoning category. This category carries a 17% weight in BenchLM.ai's overall scoring system. MuSR is currently displayed for reference but excluded from the scoring formula, so it does not directly affect overall rankings.

About MuSR

Year

2023

Tasks

Multi-step reasoning

Format

Narrative-based reasoning

Difficulty

Complex reasoning tasks

MuSR challenges models to perform multistep reasoning over complex narratives. Unlike simple factual questions, it requires models to track multiple entities, relationships, and logical steps across extended contexts.

BenchLM freshness & provenance

Version

MuSR 2023

Refresh cadence

Static

Staleness state

Stale

Question availability

Public benchmark set

StaleDisplay 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.

FAQ

What does MuSR measure?

A dataset for evaluating language models on multistep soft reasoning tasks specified in natural language narratives. Tests the ability to perform complex, structured reasoning.

Which model leads the published MuSR snapshot?

GPT-5.2 Pro currently leads the published MuSR snapshot with 95% tracked score. BenchLM shows this benchmark for display only and does not use it in overall rankings.

How many models are evaluated on MuSR?

113 AI models are included in BenchLM's mirrored MuSR snapshot, based on the public leaderboard captured on July 20, 2026.

Last updated: July 20, 2026 · mirrored from the public benchmark leaderboard

Choose a model with this week’s evidence

Join 2,000+ readers for ranking moves, pricing changes, and the claims that still need proof.

One email each week. Unsubscribe anytime.