SAGE leaderboard
As of 6 Nov 2026, the top score on SAGE is 91.3% by Claude Opus 5.5 (Anthropic), ahead of GPT-6.1 Sol at 90.5%, across 29 model configurations with a published result (Vals AI, result dated 5 Nov 2026). Images of handwritten student maths work must be graded against a rubric, requiring reading handwriting and checking reasoning. SAGE is shown for reference only and does not count toward the BenchLeader Index.
Key facts (as of 6 Nov 2026)
- Claude Opus 5.5 holds the top SAGE score at 91.3% (Vals AI, 5 Nov 2026).
- GPT-6.1 Sol is second on SAGE at 90.5% (Vals AI, 5 Nov 2026).
- SAGE: Grading handwritten maths work from images. Run by Vals AI (Vals AI).
- 29 model configurations have a published SAGE result on BenchLeader (BenchLeader, 6 Nov 2026).
- SAGE does not feed the BenchLeader Index; it is shown for reference (BenchLeader methodology).
Grading handwritten maths work from images. Run by Vals AI.
- Published by
- Vals AI
- Category
- Multimodal
- Index weight
- Reference only
- Models
- 29
- Data as of
- 6 Nov 2026
Vals AI (vals.ai).
What the test looks like
Images of handwritten student maths work must be graded against a rubric, requiring reading handwriting and checking reasoning.
How it is scored
Agreement with expert graders, run by Vals AI.
What to keep in mind
Handwriting quality and image resolution affect results.
- 1Claude Opus 5.591.3%
- 2GPT-6.1 Sol (max)90.5%
- 3Claude Haiku 5.589.8%
- 4Gemini 4 Argon (high)89.5%
- 5GLM 5.3 Flash (max)89.3%
- 6Qwen3.7 Plus89.2%
- 7Grok 4.7 (xhigh)88.7%
- 8Inkling 2 Small88.6%
- 9Gemini 3.8 Flash (high)88.2%
- 10Claude Sonnet 5.588.0%
- 11GPT-5.6 Terra (max)87.0%
- 12Kimi K3 (max)86.8%
- 13Ember-186.4%
- 14Muse Spark Open (xhigh)85.8%
- 15MiniMax M385.5%
| # | |||||
|---|---|---|---|---|---|
| 1 |  Claude Opus 5.5 | 91.3% | 63.4 | $0.283 | 2026-11-05 |
| 2 |  GPT-6.1 Solmax | 90.5% | 67.3 | $0.083 | 2026-11-05 |
| 3 | Claude Haiku 5.5 | 89.8% | – | $0.016 | 2026-11-05 |
| 4 |  Gemini 4 Argonhigh | 89.5% | 66.0 | $0.075 | 2026-11-05 |
| 5 |  GLM 5.3 Flashmaxopen ↗ | 89.3% | 54.2 | $0.0064 | 2026-11-05 |
| 6 |  Qwen3.7 Plus | 89.2% | 57.7 | $0.020 | 2026-11-05 |
| 7 |  Grok 4.7xhigh | 88.7% | 58.3 | $0.110 | 2026-11-05 |
| 8 | NewInkling 2 Small | 88.6% | 57.4 | $0.018 | 2026-11-05 |
| 9 | Gemini 3.8 Flashhigh | 88.2% | 62.0 | $0.076 | 2026-11-05 |
| 10 | Claude Sonnet 5.5 | 88.0% | – | $0.160 | 2026-11-05 |
| 11 | GPT-5.6 Terramax | 87.0% | 60.7 | $0.110 | 2026-11-05 |
| 12 |  Kimi K3maxopen ↗ | 86.8% | 62.1 | $0.227 | 2026-11-05 |
| 13 | Ember-1 | 86.4% | – | $0.205 | 2026-11-05 |
| 14 |  Muse Spark Openxhighopen ↗ | 85.8% | 54.0 | $0.069 | 2026-11-05 |
| 15 |  MiniMax M3open ↗ | 85.5% | 56.4 | $0.015 | 2026-11-05 |
| 16 | Qwen3.5 397B A17Bopen ↗ | 85.5% | 51.3 | $0.013 | 2026-11-05 |
| 17 |  DeepSeek V4.1 Flashmaxopen ↗ | 85.4% | 59.0 | $0.011 | 2026-11-05 |
| 18 | MiniMax M3.1 Flashopen ↗ | 85.3% | 53.0 | $0.0076 | 2026-11-05 |
| 19 | Qwen3.5 Flash | 84.9% | 51.8 | $0.0060 | 2026-11-05 |
| 20 |  MiMo-V2.6-Proopen ↗ | 84.5% | 57.3 | $0.021 | 2026-11-05 |
| 21 | GPT-6 Lunamax | 83.4% | 56.7 | $0.0058 | 2026-11-05 |
| 22 | MiMo-V2.6-Flashopen ↗ | 83.3% | 55.0 | $0.0074 | 2026-11-05 |
| 23 | Gemini 3.1 Prohigh | 83.2% | 60.1 | $0.054 | 2026-11-05 |
| 24 | Gemma 4 31Bhighopen ↗ | 82.8% | – | – | 2026-11-05 |
| 25 | Qwen3.8 Maxmax | 81.8% | 58.8 | $0.087 | 2026-11-05 |
| 26 | Gemini 3.5 Flash Litehigh | 79.5% | 47.9 | $0.0095 | 2026-11-05 |
| 27 | Grok 4.1 Fastthinking | 74.5% | 55.6 | $0.0047 | 2026-11-05 |
| 28 |  Mistral Medium 3.5highopen | 72.0% | – | $0.074 | 2026-11-05 |
| 29 | Mistral Small 4highopen | 59.2% | – | $0.0035 | 2026-11-05 |
More multimodal benchmarks · All LLM benchmarks
Related benchmarks
Top head-to-heads on SAGE
Cite as: BenchLeader, “SAGE leaderboard”, data as of 6 Nov 2026, https://www.benchleader.com/benchmarks/vals_sage.
SAGE: questions
- What does SAGE measure?
- Images of handwritten student maths work must be graded against a rubric, requiring reading handwriting and checking reasoning. Scores are reported in percent of tasks solved; higher is better.
- What is the top score on SAGE?
- As of 6 Nov 2026, the highest score on SAGE is 91.3% by Claude Opus 5.5 (result dated 5 Nov 2026), ahead of GPT-6.1 Sol at 90.5%. Scores come from Vals AI and are re-read daily.
- Which AI model leads SAGE?
- Claude Opus 5.5 leads SAGE with 91.3% as of 6 Nov 2026 (result dated 5 Nov 2026), ahead of GPT-6.1 Sol at 90.5%.
- How many models have SAGE results?
- 29 model configurations have a SAGE result on BenchLeader, all taken from Vals AI.
- Who runs SAGE and how often is it updated?
- SAGE is published by Vals AI. BenchLeader re-reads the published results once a day and records the date each result was published.
- Does SAGE count toward the BenchLeader Index?
- No. SAGE is shown for reference but left out of the composite index.