Share promptsSign inExplore benchmarks
← Multilingual benchmarks

MGSM leaderboard (Jan 2026): AI model scores

As of 9 Jan 2026, the highest reported Alibaba score on MGSM is Qwen 3 (235B) (Alibaba) at 92.47%, #17 on that leaderboard, per Vals AI. The source no longer updates this leaderboard.

What it measures

Maths questions in multiple languages.

Grade-school maths word problems translated into 10 languages, testing reasoning outside English.

What the score means

Accuracy

Share of problems answered correctly, averaged across languages.

Results from published AI leaderboards. Compare models within each benchmark; scores from different tests aren’t directly comparable. How we collect these results

Scores by model

Run by Vals AI (benchmark by Google Research). Leaderboard dated 9 Jan 2026 (no longer updated by the source). Last checked 22 Oct 2026. Showing top 25 of 75 models listed.

Highlight a provider: Anthropic OpenAI Google Meta DeepSeek Alibaba

Highlighting Alibaba models. Other providers are dimmed. Show all providers

  • Anthropic
  • OpenAI
  • Google
  • Meta
  • DeepSeek
  • Alibaba
MGSM: top 15 by AccuracyBars from zero. Leader: Claude Opus 4.5 (Thinking) at 95.20%. Alibaba models highlighted. Full values in the table below.0%20%40%60%80%100%Claude Opus 4.5 (Thinking) (Anthropic), rank 1: 95.20%AClaude Opus 4.5 (Thinking)95.20%Claude Opus 4.5 (Nonthinking) (Anthropic), rank 2: 94.76%AClaude Opus 4.5 (Nonthinkin…94.76%Claude Opus 4.1 (Thinking) (Anthropic), rank 3: 94.44%AClaude Opus 4.1 (Thinking)94.44%Claude Sonnet 4.5 (Thinking) (Anthropic), rank 4: 94.33%AClaude Sonnet 4.594.33%Claude Opus 4.1 (Nonthinking) (Anthropic), rank 5: 94.22%AClaude Opus 4.1 (Nonthinkin…94.22%GPT 5.2 (OpenAI), rank 6: 94.00%OGPT 5.294.00%Gemini 3 Pro (11/25) (Google), rank 7: 93.93%GGemini 3 Pro93.93%Claude Opus 4 (Nonthinking) (Anthropic), rank 8: 93.78%AClaude Opus 493.78%o4 Mini (OpenAI), rank 9: 93.42%Oo4 Mini93.42%Gemini 3 Flash (12/25) (Google), rank 10: 93.31%GGemini 3 Flash93.31%Claude Sonnet 4 (Nonthinking) (Anthropic), rank 11: 93.02%AClaude Sonnet 493.02%GPT 5.1 (OpenAI), rank 12: 92.98%OGPT 5.192.98%Claude 3.7 Sonnet (Thinking) (Anthropic), rank 13: 92.98%AClaude 3.7 Sonnet92.98%GPT 5 (OpenAI), rank 14: 92.84%OGPT 592.84%Claude 3.5 Sonnet Latest (Anthropic), rank 15: 92.58%AClaude 3.5 Sonnet Latest92.58%Qwen 3 (235B) (Alibaba), rank 17: 92.47%Q#17 Qwen 392.47%Qwen 3 Max Preview (Alibaba), rank 23: 92.15%Q#23 Qwen 3 Max Preview92.15%Qwen 3 Max (Alibaba), rank 25: 91.82%Q#25 Qwen 3 Max91.82%
MGSM: top 12 by AccuracyBars from zero. Leader: Claude Opus 4.5 (Thinking) at 95.20%. Alibaba models highlighted. Full values in the table below.0%50%100%Claude Opus 4.5 (Thinking) (Anthropic), rank 1: 95.20%AClaude Opus 4.5 …95.20%Claude Opus 4.5 (Nonthinking) (Anthropic), rank 2: 94.76%AClaude Opus 4.5 …94.76%Claude Opus 4.1 (Thinking) (Anthropic), rank 3: 94.44%AClaude Opus 4.1 …94.44%Claude Sonnet 4.5 (Thinking) (Anthropic), rank 4: 94.33%AClaude Sonnet 4.594.33%Claude Opus 4.1 (Nonthinking) (Anthropic), rank 5: 94.22%AClaude Opus 4.1 …94.22%GPT 5.2 (OpenAI), rank 6: 94.00%OGPT 5.294.00%Gemini 3 Pro (11/25) (Google), rank 7: 93.93%GGemini 3 Pro93.93%Claude Opus 4 (Nonthinking) (Anthropic), rank 8: 93.78%AClaude Opus 493.78%o4 Mini (OpenAI), rank 9: 93.42%Oo4 Mini93.42%Gemini 3 Flash (12/25) (Google), rank 10: 93.31%GGemini 3 Flash93.31%Claude Sonnet 4 (Nonthinking) (Anthropic), rank 11: 93.02%AClaude Sonnet 493.02%GPT 5.1 (OpenAI), rank 12: 92.98%OGPT 5.192.98%Qwen 3 (235B) (Alibaba), rank 17: 92.47%Q#17 Qwen 392.47%Qwen 3 Max Preview (Alibaba), rank 23: 92.15%Q#23 Qwen 3 Max P…92.15%Qwen 3 Max (Alibaba), rank 25: 91.82%Q#25 Qwen 3 Max91.82%
MGSM: top 15 by AccuracyDot plot, axis starting at 85% (not zero). Leader: Claude Opus 4.5 (Thinking) at 95.20%. Alibaba models highlighted. Full values in the table below.85%87.5%90%92.5%95%ACCURACY · AXIS STARTS AT 85%, NOT ZEROClaude Opus 4.5 (Thinking) (Anthropic), rank 1: 95.20%AClaude Opus 4.5 (Thinking)95.20%Claude Opus 4.5 (Nonthinking) (Anthropic), rank 2: 94.76%AClaude Opus 4.5 (Nonthinkin…94.76%Claude Opus 4.1 (Thinking) (Anthropic), rank 3: 94.44%AClaude Opus 4.1 (Thinking)94.44%Claude Sonnet 4.5 (Thinking) (Anthropic), rank 4: 94.33%AClaude Sonnet 4.594.33%Claude Opus 4.1 (Nonthinking) (Anthropic), rank 5: 94.22%AClaude Opus 4.1 (Nonthinkin…94.22%GPT 5.2 (OpenAI), rank 6: 94.00%OGPT 5.294.00%Gemini 3 Pro (11/25) (Google), rank 7: 93.93%GGemini 3 Pro93.93%Claude Opus 4 (Nonthinking) (Anthropic), rank 8: 93.78%AClaude Opus 493.78%o4 Mini (OpenAI), rank 9: 93.42%Oo4 Mini93.42%Gemini 3 Flash (12/25) (Google), rank 10: 93.31%GGemini 3 Flash93.31%Claude Sonnet 4 (Nonthinking) (Anthropic), rank 11: 93.02%AClaude Sonnet 493.02%GPT 5.1 (OpenAI), rank 12: 92.98%OGPT 5.192.98%Claude 3.7 Sonnet (Thinking) (Anthropic), rank 13: 92.98%AClaude 3.7 Sonnet92.98%GPT 5 (OpenAI), rank 14: 92.84%OGPT 592.84%Claude 3.5 Sonnet Latest (Anthropic), rank 15: 92.58%AClaude 3.5 Sonnet Latest92.58%Qwen 3 (235B) (Alibaba), rank 17: 92.47%Q#17 Qwen 392.47%Qwen 3 Max Preview (Alibaba), rank 23: 92.15%Q#23 Qwen 3 Max Preview92.15%Qwen 3 Max (Alibaba), rank 25: 91.82%Q#25 Qwen 3 Max91.82%
MGSM: top 12 by AccuracyDot plot, axis starting at 85% (not zero). Leader: Claude Opus 4.5 (Thinking) at 95.20%. Alibaba models highlighted. Full values in the table below.85%90%95%100%AXIS STARTS AT 85%, NOT ZEROClaude Opus 4.5 (Thinking) (Anthropic), rank 1: 95.20%AClaude Opus 4.5 …95.20%Claude Opus 4.5 (Nonthinking) (Anthropic), rank 2: 94.76%AClaude Opus 4.5 …94.76%Claude Opus 4.1 (Thinking) (Anthropic), rank 3: 94.44%AClaude Opus 4.1 …94.44%Claude Sonnet 4.5 (Thinking) (Anthropic), rank 4: 94.33%AClaude Sonnet 4.594.33%Claude Opus 4.1 (Nonthinking) (Anthropic), rank 5: 94.22%AClaude Opus 4.1 …94.22%GPT 5.2 (OpenAI), rank 6: 94.00%OGPT 5.294.00%Gemini 3 Pro (11/25) (Google), rank 7: 93.93%GGemini 3 Pro93.93%Claude Opus 4 (Nonthinking) (Anthropic), rank 8: 93.78%AClaude Opus 493.78%o4 Mini (OpenAI), rank 9: 93.42%Oo4 Mini93.42%Gemini 3 Flash (12/25) (Google), rank 10: 93.31%GGemini 3 Flash93.31%Claude Sonnet 4 (Nonthinking) (Anthropic), rank 11: 93.02%AClaude Sonnet 493.02%GPT 5.1 (OpenAI), rank 12: 92.98%OGPT 5.192.98%Qwen 3 (235B) (Alibaba), rank 17: 92.47%Q#17 Qwen 392.47%Qwen 3 Max Preview (Alibaba), rank 23: 92.15%Q#23 Qwen 3 Max P…92.15%Qwen 3 Max (Alibaba), rank 25: 91.82%Q#25 Qwen 3 Max91.82%
Accuracy
Share of problems answered correctly, averaged across languages.

Last updated by Vals in January 2026, so newer models are missing.

All scores

MGSM scores, Alibaba models highlighted
RankModelProviderAccuracyPrice in / out ($/1M tokens)
1 Claude Opus 4.5 (Thinking)Anthropic95.20%$5 / $25
2 Claude Opus 4.5 (Nonthinking)Anthropic94.76%$5 / $25
3 Claude Opus 4.1 (Thinking)Anthropic94.44%$15 / $75
4 Claude Sonnet 4.5 (Thinking)Anthropic94.33%$3 / $15
5 Claude Opus 4.1 (Nonthinking)Anthropic94.22%$15 / $75
6 GPT 5.2OpenAI94.00%$1.75 / $14
7 Gemini 3 Pro (11/25)Google93.93%$2 / $12
8 Claude Opus 4 (Nonthinking)Anthropic93.78%$15 / $75
9 o4 MiniOpenAI93.42%$1.1 / $4.4
10 Gemini 3 Flash (12/25)Google93.31%$0.5 / $3
Show 15 more rows
RankModelProviderAccuracyPrice in / out ($/1M tokens)
11 Claude Sonnet 4 (Nonthinking)Anthropic93.02%$3 / $15
12 GPT 5.1OpenAI92.98%$1.25 / $10
13 Claude 3.7 Sonnet (Thinking)Anthropic92.98%$3 / $15
14 GPT 5OpenAI92.84%$1.25 / $10
15 Claude 3.5 Sonnet LatestAnthropic92.58%$3 / $15
16 GPT 5 MiniOpenAI92.58%$0.25 / $2
17 Qwen 3 (235B)Alibaba92.47%$0.22 / $0.88
18 Llama 4 MaverickMeta92.44%$0.22 / $0.88
19 Claude 3.7 Sonnet (Nonthinking)Anthropic92.40%$3 / $15
20 DeepSeek R1DeepSeek92.25%$1.35 / $5.4
21 Claude Haiku 4.5 (Thinking)Anthropic92.15%$1 / $5
22 DeepSeek V3DeepSeek92.15%$0.9 / $0.9
23 Qwen 3 Max PreviewAlibaba92.15%$1.2 / $6
24 GPT OSS 120BOpenAI92.04%$0.15 / $0.6
25 Qwen 3 MaxAlibaba91.82%$1.2 / $6

Source: Vals AI leaderboard ↗

Best-ranked Alibaba model here: Qwen 3 (235B) (Alibaba). Browse prompts for any model →