SAGE (Vals AI) leaderboard
- Claude Opus 4.7 leads
- 56.1%
- models measured
- 74
- lead over #2
- +1.1 pts
- median
- 45.7%
Leader on SAGE (Vals AI) · read October 21
On October 21, 2026, 74 of our catalog's models had a SAGE (Vals AI) result, and Anthropic Claude Opus 4.7 led with 56.1%, ahead of Google Gemma 4 31B at 55.0%.
Grading students' math work against a rubric. The accuracy Vals AI publishes for it.
Role in the index
SAGE (Vals AI) is shown as published and does not feed any of our indexes. Read the methodology
| # | Model | Result | Latency, as printed | Evidence | Coding index | Output $/Mtok |
|---|---|---|---|---|---|---|
| 1 | Claude Opus 4.7Anthropic | 56.1% | 2m 05s | 67.5 | $25.00 | |
| 2 | Gemma 4 31BGoogle | 55.0% | 1m 33s | 39.0 | $0.40 | |
| 3 | Claude Opus 4.8Anthropic | 54.8% | 3m 06s | 65.0 | $25.00 | |
| 4 | Gemini 4 ArgonGoogle | 53.6% | 1m 03s | 81.0 | $10.00 | |
| 5 | Kimi K3Moonshot AI | 52.8% | 47 s | 62.3 | $15.00 | |
| 6 | GPT-5.6 SolOpenAI | 52.6% | 1m 15s | 73.5 | $20.00 | |
| 7 | Qwen3.8 27BQwen | 52.4% | 1m 23s | 74.2 | $2.55 | |
| 8 | Claude Opus 4.5Anthropic | 52.1% | 1m 27s | 44.0 | $25.00 | |
| 9 | Claude Fable 5Anthropic | 51.9% | 1m 57s | 79.3 | $50.00 | |
| 10 | Gemini 3 Flash PreviewGoogle | 51.8% | 43 s | 43.7 | $3.00 | |
| 11 | Claude Sonnet 5.5Anthropic | 51.8% | 1m 53s | 87.3 | $10.00 | |
| 12 | Claude Opus 4.6Anthropic | 51.6% | 2m 41s | 57.2 | $25.00 | |
| 13 | GPT-5.5OpenAI | 51.5% | 1m 16s | 71.3 | $30.00 | |
| 14 | Qwen3.8 Max (0902)Qwen | 51.2% | 2m 13s | 62.4 | $6.00 | |
| 15 | GPT-5.4 MiniOpenAI | 50.8% | 2m 21s | 35.8 | $4.50 | |
| 16 | MiniMax M3MiniMax | 50.6% | 1m 59s | 39.2 | $1.20 | |
| 17 | Kimi K2.6Moonshot AI | 50.2% | 4m 31s | 52.1 | $4.00 | |
| 18 | Gemini 3.5 FlashGoogle | 49.9% | 33 s | 39.8 | $9.00 | |
| 19 | Kimi K2.5Moonshot AI | 49.9% | 2m 53s | 38.6 | $2.85 | |
| 20 | Gemini 3.1 Flash Lite PreviewGoogle | 49.5% | 16 s | 22.7 | $1.50 | |
| 21 | Claude Opus 5Anthropic | 49.4% | 56 s | 83.0 | $25.00 | |
| 22 | GPT-5.2OpenAI | 49.3% | 2m 59s | 55.9 | $14.00 | |
| 23 | Gemini 3.7 FlashGoogle | 49.2% | 9.8 s | 58.0 | $3.75 | |
| 24 | Claude Sonnet 5Anthropic | 48.9% | 7m 14s | 57.7 | $10.00 | |
| 25 | Gemini 3.6 FlashGoogle | 48.7% | 22 s | 62.5 | $3.75 | |
| 26 | Gemini 3.1 Pro PreviewGoogle | 48.7% | 1m 01s | 32.3 | $12.00 | |
| 27 | Claude Fable 5.1Anthropic | 48.5% | 1m 46s | 84.2 | $50.00 | |
| 28 | GPT-6 LunaOpenAI | 48.1% | 1m 31s | 57.3 | $0.50 | |
| 29 | DeepSeek V4.1 FlashDeepSeek | 47.9% | 36 s | 62.5 | $0.78 | |
| 30 | Muse Spark 1.2Meta | 47.7% | 29 s | 55.7 | $4.25 | |
| 31 | Gemini 3.5 Flash LiteGoogle | 47.3% | 11 s | 49.0 | $2.50 | |
| 32 | GPT-5.6 TerraOpenAI | 47.0% | 40 s | 63.9 | $12.00 | |
| 33 | Claude Sonnet 4.6Anthropic | 46.6% | 2m 43s | 57.3 | $15.00 | |
| 34 | GPT-6.1 SolOpenAI | 46.5% | 50 s | 82.4 | $10.00 | |
| 35 | GPT-6 AstraOpenAI | 46.4% | 42 s | 85.1 | $50.00 | |
| 36 | Claude Opus 5.5Anthropic | 45.8% | 1m 54s | 88.8 | $20.00 | |
| 37 | Claude Haiku 5.5Anthropic | 45.8% | 2m 23s | 70.7 | $0.50 | |
| 38 | Muse Spark 1.1Meta | 45.6% | 50 s | 70.3 | $4.25 | |
| 39 | Qwen3.6 27BQwen | 45.6% | 2m 39s | 36.8 | $2.70 | |
| 40 | MiMo-V2.6-ProXiaomi | 45.0% | 3m 40s | 65.4 | $0.87 | |
| 41 | Qwen3.6 PlusQwen | 44.9% | 3m 49s | 44.1 | $1.95 | |
| 42 | GPT-6 SolOpenAI | 44.8% | 31 s | 76.0 | $10.00 | |
| 43 | Gemini 2.5 FlashGoogle | 44.8% | 40 s | Not benchmarked | $2.50 | |
| 44 | GPT-5.6 LunaOpenAI | 44.2% | 52 s | 58.0 | $1.20 | |
| 45 | GPT-5OpenAI | 43.7% | 59 s | 38.8 | $10.00 | |
| 46 | MiMo-V2.6-FlashXiaomi | 43.5% | 2m 16s | 54.7 | $0.28 | |
| 47 | GPT-5.4OpenAI | 43.3% | 2m 32s | 64.4 | $15.00 | |
| 48 | MiMo-V2.5Xiaomi | 43.3% | 53 s | 50.0 | $0.336 | |
| 49 | GPT-5.1OpenAI | 43.2% | 1m 38s | 40.4 | $10.00 | |
| 50 | GPT-5 MiniOpenAI | 43.0% | 37 s | 34.2 | $2.00 | |
| 51 | Qwen3.5-FlashQwen | 42.5% | 1m 25s | 41.1 | $0.26 | |
| 52 | Gemini 2.5 ProGoogle | 41.9% | 53 s | 19.1 | $10.00 | |
| 53 | o3OpenAI | 41.8% | 48 s | Not benchmarked | $8.00 | |
| 54 | o4 MiniOpenAI | 41.1% | 27 s | Not benchmarked | $4.40 | |
| 55 | Grok 4.7xAI | 40.8% | 4m 51s | 66.7 | $6.00 | |
| 56 | Qwen3.7 PlusQwen | 39.3% | 2m 57s | 33.2 | $1.28 | |
| 57 | Grok 4.20xAI | 38.2% | 1m 06s | 33.3 | $2.50 | |
| 58 | GPT-5.4 NanoOpenAI | 38.1% | 34 s | 44.8 | $1.25 | |
| 59 | Mistral Medium 3.5Mistral AI | 37.6% | 56 s | 25.5 | $7.50 | |
| 60 | T InklingThinking Machines | 36.5% | 3m 18s | 31.5 | $4.05 | |
| 61 | Claude Sonnet 4.5Anthropic | 36.1% | 3m 30s | 39.5 | $15.00 | |
| 62 | Llama 4 MaverickMeta Llama | 35.7% | 44 s | Not benchmarked | $1.00 | |
| 63 | Gemini 3.8 FlashGoogle | 35.1% | 29 s | 61.6 | $3.75 | |
| 64 | Grok 4.5xAI | 35.0% | 2m 50s | 58.6 | $6.00 | |
| 65 | Llama 4 ScoutMeta Llama | 34.8% | 33 s | Not benchmarked | $0.50 | |
| 66 | T Inkling SmallThinking Machines | 32.7% | 3m 26s | 27.3 | $1.20 | |
| 67 | Claude Haiku 4.5Anthropic | 31.8% | 1m 06s | 35.6 | $5.00 | |
| 68 | Claude Opus 4.1Anthropic | 31.4% | 17 s | Not benchmarked | $75.00 | |
| 69 | Gemini 2.5 Flash LiteGoogle | 30.8% | 33 s | Not benchmarked | $0.40 | |
| 70 | Qwen3.5 Plus 2026-04-20Qwen | 30.4% | 5m 13s | 39.5 | $1.80 | |
| 71 | GPT-5 NanoOpenAI | 30.4% | 39 s | Not benchmarked | $0.40 | |
| 72 | Grok 4.6xAI | 28.9% | 3m 22s | 65.7 | $6.00 | |
| 73 | Mistral Large 3 2512Mistral AI | 24.6% | 55 s | Not benchmarked | $1.50 | |
| 74 | Grok 4.3xAI | 19.7% | 1m 36s | 37.4 | $2.50 |
Only models in our catalog are listed, and each appears once with its best published run. 136 rows were read from the source on October 21, 2026, and rows that match no catalog model are left out. Every other column is a figure the board publishes for the same run.