MGSM
AI benchmark results and model performance
Type: percentageMath
Recorded results
40
Average score
79.07%
Score range
47.90% - 92.30%
Leading model
92.30% - Llama 4 Maverick
Scores over time
Individual benchmark scores plotted by date.
Model results
| Organisation | Models | Reported | Top score | Info | Self reported | Source |
|---|---|---|---|---|---|---|
| Llama 4 Maverick | 05 Apr 2025 | 92.30% | - | Yes | Source | |
| o3 mini | 30 Jan 2025 | 92% | - | Yes | Source | |
| Llama 2 70B Chat | 20 Jun 2023 | 91.10% | inferred family alias from llama-3.3-70b-instruct (score=0.3129; benches=9) | Yes | Source | |
| o1 preview | 12 Sept 2024 | 90.80% | - | Yes | Source | |
| Llama 4 Scout | 05 Apr 2025 | 90.60% | - | Yes | Source | |
| GPT 4o (2024-05-13) | 13 May 2024 | 90.50% | - | Yes | Source | |
| GPT 4 Turbo Preview (1106) | 06 Nov 2023 | 88.50% | inferred version-family alias from gpt-4-turbo-2024-04-09 | Yes | Source | |
| Gemini 1.5 Pro 001 | 23 May 2024 | 87.50% | inferred alias from gemini-1.5-pro | Yes | Source | |
| Gemini 1.5 Pro Exp (2024-08-01) | 01 Aug 2024 | 87.50% | inferred alias from gemini-1.5-pro | Yes | Source | |
| Gemini 1.5 Pro Exp (2024-08-27) | 27 Aug 2024 | 87.50% | inferred alias from gemini-1.5-pro | Yes | Source | |
| Gemini Robotics ER 1.5 Preview | 25 Sept 2025 | 87.50% | inferred family alias from gemini-1.5-pro (score=0.3717; benches=23) | Yes | Source | |
| LearnLM 1.5 Pro Experimental | 19 Nov 2024 | 87.50% | inferred family alias from gemini-1.5-pro (score=0.3700; benches=23) | Yes | Source | |
| GPT 4o Mini Audio Preview | 17 Dec 2024 | 87% | inferred modality/version alias from gpt-4o-mini-2024-07-18 | Yes | Source | |
| GPT 4o Mini Realtime Preview | 17 Dec 2024 | 87% | inferred modality/version alias from gpt-4o-mini-2024-07-18 | Yes | Source | |
| GPT 4o Mini Search Preview | 11 Mar 2025 | 87% | inferred modality/version alias from gpt-4o-mini-2024-07-18 | Yes | Source | |
| GPT 4o Mini Transcribe (2025-12-15) | 15 Dec 2025 | 87% | inferred modality/version alias from gpt-4o-mini-2024-07-18 | Yes | Source | |
| GPT 4o Mini Transcribe (2025-03-20) | 20 Mar 2025 | 87% | inferred modality/version alias from gpt-4o-mini-2024-07-18 | Yes | Source | |
| GPT 4o Mini TTS (2025-12-15) | 15 Dec 2025 | 87% | inferred modality/version alias from gpt-4o-mini-2024-07-18 | Yes | Source | |
| GPT 4o Mini TTS (2025-03-20) | 20 Mar 2025 | 87% | inferred modality/version alias from gpt-4o-mini-2024-07-18 | Yes | Source | |
| Llama 3.2 90B Vision Instruct | - | 86.90% | inferred high-confidence family alias from llama-3.2-90b-instruct (score=0.5873; benches=13) | Yes | Source | |
| Qwen 3 235B A22B | - | 83.53% | - | Yes | Source | |
| Gemini 1.5 Flash 001 | 23 May 2024 | 82.60% | inferred alias from gemini-1.5-flash | Yes | Source | |
| Gemini 1.5 Flash Preview | 14 May 2024 | 82.60% | inferred alias from gemini-1.5-flash | Yes | Source | |
| Granite 4.1 8B | 29 Apr 2026 | 82.32% | 8-shot | Yes | Source | |
| Phi 1 | - | 80.60% | inferred family alias from phi-4 (score=0.3100; benches=13) | Yes | Source |
