Skip to main content
BenchLM

Vals MGSM (MGSM)

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

Data verified 36 confirmed releases in the last 30 daysFollow model changes

A multilingual benchmark for mathematical questions.

MGSM score on MGSM — January 9, 2026

We mirror the published mgsm score view for MGSM. Claude Opus 4.5 Thinking leads the public snapshot at 95.20%, followed by Claude Opus 4.5 (94.76%) and Claude Opus 4.1 20250805 Thinking (94.44%). We do not use these results to rank models overall.

75 modelsMathematicsCurrentDisplay onlyUpdated January 9, 2026

MGSM score table (75 models)

Score
1
Claude Opus 4.5 ThinkingAnthropic · Closed
95.20%
2
Claude Opus 4.5Anthropic · Closed
94.76%
4
Claude Sonnet 4.5 ThinkingAnthropic · Closed
94.33%
5
94.22%
6
GPT-5.2OpenAI · Closedxhigh reasoning
94.00%
7
Gemini 3 Pro PreviewGooglehigh reasoning
93.93%
8
Claude Opus 4Anthropic
93.78%
9
O4 MiniOpenAIhigh reasoning
93.42%
10
Gemini 3 Flash PreviewGooglehigh reasoning
93.31%
11
93.02%
12
GPT-5.1OpenAI · Closedhigh reasoning
92.98%
14
GPT-5OpenAIhigh reasoning
92.84%
15
Claude 3.5 SonnetAnthropic · Closed
92.58%
16
GPT-5 miniOpenAI · Closedhigh reasoning
92.58%
17
Qwen3 235b A22bFireworks AI
92.47%
19
92.40%
20
DeepSeek-R1DeepSeek · Open weight
92.25%
21
Claude Haiku 4.5 ThinkingAnthropic · Closed
92.15%
22
DeepSeek V3DeepSeek · Open weight
92.15%
23
92.15%
24
GPT-OSS 120BOpenAI · Open weight
92.04%
25
Qwen3 MaxAlibaba · Closed
91.82%
26
o3OpenAI · Closedhigh reasoning
91.75%
27
DeepSeek V3 0324Fireworks AI
91.67%
28
91.35%
29
o3-miniOpenAI · Closedhigh reasoning
91.35%
31
90.95%
32
90.91%
33
90.87%
34
90.87%
35
DeepSeek V3p2Fireworks
90.87%
37
GLM-4.5Z.AI · Closed
90.84%
38
GPT-4oOpenAI · Closedhigh reasoning
90.69%
39
90.44%
40
GPT-4oOpenAI · Closedhigh reasoning
90.36%
41
90.36%
42
Kimi K2 ThinkingMoonshot AI
90.15%
45
GLM-4.6Z.AI · Open weight
89.75%
47
89.53%
48
o1OpenAI · Closedhigh reasoning
89.31%
49
GPT-5 nanoOpenAI · Closedhigh reasoning
89.31%
50
89.20%
51
89.02%
52
GPT-OSS 20BOpenAI · Open weight
89.02%
54
GLM-4.7Z.AI · Open weight
88.18%
57
87.85%
58
GPT-4.1 miniOpenAI · Closedhigh reasoning
87.78%
59
GPT-4.1OpenAI · Closedhigh reasoning
87.67%
61
87.24%
62
86.58%
63
86.25%
64
GPT-4o miniOpenAI · Closedhigh reasoning
86.18%
65
86.15%
66
86.04%
67
85.71%
68
85.42%
69
84.62%
70
84.22%
71
83.96%
72
74.62%
73
71.24%
74
GPT-4.1 nanoOpenAI · Closedhigh reasoning
69.27%
75
41.71%

How MGSM is shown here

BenchLM mirrors the public Vals AI MGSM leaderboard captured from https://www.vals.ai/benchmarks/mgsm and updated by Vals on January 9, 2026. The snapshot preserves overall scores, uncertainty, latency, cost-per-test metadata, and task-level scores where Vals publishes them.

MGSM is display only on BenchLM. Vals proprietary or Vals-hosted aggregate views are useful context, but BenchLM does not use them as weighted ranking inputs or as a replacement for benchmark-native source records.

Snapshot

75 Vals rows12 task viewspublic datasetTasks: Overall, English, Japanese, Chinese, BengaliDisplay only

The published MGSM snapshot places Claude Opus 4.5 Thinking first at 95.20%. The third row is 0.76 points behind. The broader top-10 range is 1.89 points, so many of the published results sit in a relatively narrow band.

75 models have been evaluated on MGSM. The benchmark falls in the Mathematics category. We keep external benchmark mirrors separate from the weighted global scoring system, so these results remain source-specific evidence. MGSM is currently displayed for reference but excluded from the scoring formula, so it does not directly affect overall rankings.

About MGSM

Year

2026

Tasks

Multilingual grade-school math questions

Format

Accuracy score

Difficulty

Multilingual mathematical reasoning

BenchLM mirrors the public Vals AI MGSM leaderboard as display-only external evidence. The captured snapshot preserves overall scores, task-level scores where Vals publishes them, uncertainty, latency, and cost-per-test metadata. It is excluded from BenchLM weighted rankings.

Freshness and provenance

Version

MGSM 2026

Refresh cadence

Quarterly

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 MGSM measure?

A multilingual benchmark for mathematical questions.

Which model leads the published MGSM snapshot?

Claude Opus 4.5 Thinking currently leads the published MGSM snapshot with 95.20% mgsm score. BenchLM shows this benchmark for display only and does not use it in overall rankings.

How many models are evaluated on MGSM?

The January 9, 2026 snapshot contains 75 AI models.

Last updated: January 9, 2026 · 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 2,000+ readers.

One email each week. Unsubscribe anytime.