SAGE (Service Agent) - Online Education: leaderboard

SAGE overall assessment score in the online education customer-service scenario: online education contract disputes and refund-risk control, combining procedure compliance and conversational quality on a 0-100 scale, higher is better. It is unrelated to the SAGE robustness benchmark.

Metric: Overall Assessment Score (0-100), 0.8 x logical compliance + 0.2 conversational quality, in the Online Education scenario (online education contract disputes and refund-risk control), over SAGE customer-service dialogues: standard operating procedures formalized as dialogue graphs, an adversarial simulated user with varied intents and personas, turns 1, 5, 10, 15 and the final turn scored; three unnamed judge LLMs (majority vote and mean) plus a deterministic rule engine produce the reference; closed models through their APIs, open models served with vLLM; 0-100 scale; higher is better. Source: arxiv.org. Saturation forecast: Around February 2027. 27 models tracked. Updated 7 Oct 2026.

Top models

#ModelScoreOverall rank
1Gemini 3 Pro (Preview)82.15#119
2DeepSeek V3.278.17#419
3Gemini 3 Flash (Preview)75.62#152
4Claude Opus 4.575.22#160
5Qwen Max (SAGE checkpoint unspecified)74.75
6Claude Sonnet 4.574.24#261
7MiniMax-M2.172.55#575
8GLM-4.772.36#380
9GPT-4.171.69#505
10Kimi K2.571.15#266
11Gemini 2.5 Pro70.28#276
12DeepSeek R169.4#482
13DeepSeek V368.77#729
14Qwen 2.5 72B Instruct68.76#1117
15Qwen 3 235B A22B68.02#628

No result here: #1 Claude Fable 5.2, #3 Gemini 4 Neon, #4 GPT-6.2 Astra.

Interactive version: theaggregate.ai/benchmark?slug=sage-service-agent-online-education · How It Works · Data refreshed daily, snapshot 2026-11-21.