Vals SAGE (SAGE)
We show this table for reference; we do not rank on it.
Student Assessment with Generative Evaluation
Evaluation results published by Vals AI. Published leaderboard aggregates and evaluation settings only. Dataset contents are not included.
SAGE score on SAGE — October 8, 2026
We mirror the published sage score view for SAGE. Claude Opus 4.7 leads the public snapshot at 56.10%, followed by Gemma 4 31b It (55.03%) and Claude Opus 4.8 (54.79%). We do not use these results to rank models overall.
Claude Opus 4.7
Anthropic
Gemma 4 31b It
high reasoning
Claude Opus 4.8
Anthropic
91 modelsMultimodal & GroundedCurrentDisplay onlyUpdated October 8, 2026
SAGE score table (91 models)
| Rank | Model / configuration | Score | Parameters (B) | Open / closed |
|---|---|---|---|---|
| 1 | Claude Opus 4.7Anthropic | 56.10% | Not reported | Closed |
| 2 | Gemma 4 31b ItGooglehigh reasoning | 55.03% | Not reported | Unknown |
| 3 | Claude Opus 4.8Anthropic | 54.79% | Not reported | Closed |
| 4 | Gemini 4 ArgonGooglehigh reasoning | 53.65% | Not reported | Closed |
| 5 | Kimi K3Moonshot AImax reasoning | 52.78% | Not reported | Pending |
| 6 | GPT-5.6 SolOpenAImax reasoning | 52.56% | Not reported | Closed |
| 7 | Qwen3.8-27BAlibabaxhigh reasoning | 52.40% | Not reported | Open |
| 8 | Claude Opus 4.5 ThinkingAnthropic | 52.09% | Not reported | Closed |
| 9 | Claude Fable 5Anthropic | 51.89% | Not reported | Closed |
| 10 | Gemini 3 Flash PreviewGooglehigh reasoning | 51.85% | Not reported | Unknown |
| 11 | Claude Sonnet 5.5Anthropic | 51.77% | Not reported | Closed |
| 12 | Claude Opus 4.6 (Adaptive)Anthropic | 51.58% | Not reported | Closed |
| 13 | GPT-5.5OpenAIxhigh reasoning | 51.53% | Not reported | Closed |
| 14 | Qwen3.8 MaxAlibaba | 51.25% | Not reported | Open |
| 15 | GPT-5.4 miniOpenAIxhigh reasoning | 50.81% | Not reported | Closed |
| 16 | MiniMax M3MiniMax | 50.57% | Not reported | Open |
| 17 | Kimi K2.6Moonshot AI | 50.22% | Not reported | Open |
| 18 | Gemini 3.5 FlashGooglehigh reasoning | 49.88% | Not reported | Closed |
| 19 | Kimi K2.5 ThinkingMoonshot AI | 49.87% | Not reported | Unknown |
| 20 | Gemini 3.1 Flash Lite PreviewGooglehigh reasoning | 49.54% | Not reported | Unknown |
| 21 | Claude Opus 5Anthropic | 49.43% | Not reported | Closed |
| 22 | GPT-5.2OpenAIxhigh reasoning | 49.27% | Not reported | Closed |
| 23 | Gemini 3.7 FlashGooglehigh reasoning | 49.23% | Not reported | Closed |
| 24 | Claude Sonnet 5Anthropic | 48.92% | Not reported | Closed |
| 25 | Gemini 3.6 FlashGooglehigh reasoning | 48.69% | Not reported | Closed |
| 26 | Gemini 3.1 Pro PreviewGooglehigh reasoning | 48.68% | Not reported | Unknown |
| 27 | Claude Fable 5.1Anthropic | 48.53% | Not reported | Closed |
| 28 | GPT-6 LunaOpenAImax reasoning | 48.09% | Not reported | Closed |
| 29 | DeepSeek V4.1 Flash 1017DeepSeekhigh reasoning | 47.88% | Not reported | Open |
| 30 | Muse Spark 1.2Metaxhigh reasoning | 47.66% | Not reported | Closed |
| 31 | Gemini 3 Pro PreviewGooglehigh reasoning | 47.62% | Not reported | Unknown |
| 32 | Gemini 3.5 Flash-LiteGooglehigh reasoning | 47.30% | Not reported | Closed |
| 33 | GPT-5.6 TerraOpenAImax reasoning | 47.00% | Not reported | Closed |
| 34 | Claude Sonnet 4.6Anthropic | 46.58% | Not reported | Closed |
| 35 | GPT-6.1 SolOpenAImax reasoning | 46.54% | Not reported | Closed |
| 36 | GPT-6 AstraOpenAImax reasoning | 46.37% | Not reported | Closed |
| 37 | Claude Opus 5.5Anthropic | 45.83% | Not reported | Closed |
| 38 | Claude Haiku 5.5Anthropic | 45.78% | Not reported | Closed |
| 39 | Muse Spark 1.1Metaxhigh reasoning | 45.58% | Not reported | Closed |
| 40 | Qwen3.6-27BAlibaba | 45.58% | Not reported | Open |
| 41 | MiMo-V2.6-ProXiaomi | 45.05% | Not reported | Open |
| 42 | Claude Opus 4.5Anthropic | 45.00% | Not reported | Closed |
| 43 | Qwen3.6 PlusAlibaba | 44.86% | Not reported | Closed |
| 44 | GPT-6 SolOpenAImax reasoning | 44.79% | Not reported | Closed |
| 45 | Gemini 2.5 FlashGoogle | 44.76% | Not reported | Closed |
| 46 | GPT-5.6 LunaOpenAImax reasoning | 44.22% | Not reported | Closed |
| 47 | Gemini 2.5 Flash Preview 09 2025Google | 44.11% | Not reported | Unknown |
| 48 | GPT-5OpenAIhigh reasoning | 43.68% | Not reported | Unknown |
| 49 | MiMo-V2.6-FlashXiaomi | 43.53% | Not reported | Open |
| 50 | GPT-5.4OpenAIxhigh reasoning | 43.31% | Not reported | Closed |
| 51 | MiMo-V2.5Xiaomi | 43.27% | Not reported | Closed |
| 52 | GPT-5.1OpenAIhigh reasoning | 43.23% | Not reported | Closed |
| 53 | GPT-5 miniOpenAIhigh reasoning | 42.99% | Not reported | Closed |
| 54 | Qwen3.5 FlashAlibaba | 42.49% | Not reported | Closed |
| 55 | Qwen3 VL PlusAlibaba | 42.10% | Not reported | Unknown |
| 56 | Gemini 2.5 ProGoogle | 41.92% | Not reported | Closed |
| 57 | o3OpenAIhigh reasoning | 41.77% | Not reported | Closed |
| 58 | o4-miniOpenAIhigh reasoning | 41.06% | Not reported | Closed |
| 59 | Grok 4.7xAIxhigh reasoning | 40.79% | Not reported | Closed |
| 60 | Gemini 2.5 Flash ThinkingGoogle | 39.60% | Not reported | Unknown |
| 61 | Gemini 2.5 Flash Preview 09 2025 ThinkingGoogle | 39.38% | Not reported | Unknown |
| 62 | Qwen3.7 PlusAlibaba | 39.25% | Not reported | Closed |
| 63 | Grok 4.20 0309 ReasoningxAI | 38.24% | Not reported | Unknown |
| 64 | GPT-5.4 nanoOpenAIhigh reasoning | 38.08% | Not reported | Closed |
| 65 | Mistral Medium 3.5Mistral AIhigh reasoning | 37.61% | Not reported | Unknown |
| 66 | InklingThinking Machines Lab0.99 reasoning | 36.55% | Not reported | Open |
| 67 | Claude Sonnet 4.5 ThinkingAnthropic | 36.06% | Not reported | Closed |
| 68 | Llama4 Maverick Instruct BasicFireworks AI | 35.70% | Not reported | Unknown |
| 69 | Gemini 3.8 FlashGooglehigh reasoning | 35.06% | Not reported | Closed |
| 70 | Claude Sonnet 4Anthropic | 35.00% | Not reported | Unknown |
| 71 | Grok 4.5xAIhigh reasoning | 34.97% | Not reported | Closed |
| 72 | Meta Llama Llama 4 Scout 17B 16E InstructTogether AI | 34.83% | Not reported | Unknown |
| 73 | Claude Sonnet 4.5Anthropic | 32.88% | Not reported | Closed |
| 74 | Inkling-SmallThinking Machines Lab0.99 reasoning | 32.74% | Not reported | Open |
| 75 | Claude Sonnet 4 20250514 ThinkingAnthropic | 32.01% | Not reported | Unknown |
| 76 | Claude Haiku 4.5 ThinkingAnthropic | 31.82% | Not reported | Closed |
| 77 | Claude Opus 4.1Anthropic | 31.38% | Not reported | Unknown |
| 78 | Grok 4.1 Fast (Reasoning)xAI | 31.33% | Not reported | Closed |
| 79 | Gemini 2.5 Flash Lite Preview 09 2025 ThinkingGoogle | 30.78% | Not reported | Unknown |
| 80 | Qwen3.5 Plus ThinkingAlibaba | 30.40% | Not reported | Unknown |
| 81 | Claude Opus 4.1 20250805 ThinkingAnthropic | 30.39% | Not reported | Unknown |
| 82 | GPT-5 nanoOpenAIhigh reasoning | 30.38% | Not reported | Closed |
| 83 | Grok 4 Fast (Reasoning)xAI | 29.76% | Not reported | Closed |
| 84 | Grok 4.6xAIhigh reasoning | 28.90% | Not reported | Closed |
| 85 | Gemini 2.5 Flash Lite Preview 09 2025Google | 27.97% | Not reported | Unknown |
| 86 | Gemini 2.5 Flash-LiteGoogle | 27.53% | Not reported | Closed |
| 87 | Grok 4 0709xAI | 25.10% | Not reported | Unknown |
| 88 | Mistral Large 2512Mistral AI | 24.59% | Not reported | Unknown |
| 89 | Grok 4.3xAI | 19.74% | Not reported | Closed |
| 90 | Grok 4 Fast Non ReasoningxAI | 15.99% | Not reported | Unknown |
| 91 | Grok 4.1 Fast Non ReasoningxAI | 11.91% | Not reported | Unknown |
How SAGE is shown here
BenchLM mirrors the public Vals AI SAGE leaderboard captured from https://www.vals.ai/benchmarks/sage and updated by Vals on October 8, 2026. The snapshot preserves overall scores, uncertainty, latency, cost-per-test metadata, and task-level scores where Vals publishes them.
SAGE 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
The published SAGE snapshot places Claude Opus 4.7 first at 56.10%. The third row is 1.32 points behind. The broader top-10 range is 4.25 points, so many of the published results sit in a relatively narrow band.
91 models have been evaluated on SAGE. The benchmark falls in the Multimodal & Grounded category. We keep external benchmark mirrors separate from the weighted global scoring system, so these results remain source-specific evidence. SAGE is currently displayed for reference but excluded from the scoring formula, so it does not directly affect overall rankings.
About SAGE
Year
2026
Tasks
Student assessment with generative evaluation
Format
Accuracy score
Difficulty
Education assessment reasoning
BenchLM mirrors the public Vals AI SAGE 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
SAGE 2026
Refresh cadence
Quarterly
Staleness state
Current
Question availability
Public benchmark set
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 SAGE measure?
Student Assessment with Generative Evaluation
Which model leads the published SAGE snapshot?
Claude Opus 4.7 currently leads the published SAGE snapshot with 56.10% sage score. BenchLM shows this benchmark for display only and does not use it in overall rankings.
How many models are evaluated on SAGE?
The October 8, 2026 snapshot contains 91 AI models.
Know when it’s worth switching models
The model to choose, the cheaper alternative, and the release we would wait on.
Read a sample issueJoin 5,600+ readers.
One email each week. Unsubscribe anytime.