The most energy-efficient tested configuration consumes 1.8 mWh per query, a 43.8% reduction relative to the Qwen-7B FP16 baseline¶

Claim · All claims
Evidence · 1 study · design r2 · q2
Currency · newest study 2026 · not yet checked against its sources · not reviewed by a person

Subclaims¶

q2 i? DeepSeek-MoE (Ours) consumes 1.8 mWh per query (1.8 Wh per 1,000 queries), 43.8% less than the Qwen-7B FP16 baseline of 3.2 mWh/query. → Shen X 2026

Evidence¶

Shen X 2026¶

Shen X, Feng L, Hua S, Liu D, Xie Z and Liu B (2026) Towards sustainable AI knowledge-base assistants in computer science education: on-premise deployment and optimization with open educational resources. Front. Psychol. 17:1843444. https://doi.org/10.3389/fpsyg.2026.1843444

q2 · i? · design · r2

Energy efficiency analysis (Section 5.3) with GPU-level measurements at the device rail, idle power subtracted, 10 Hz sampling. The article reports "1.8 mWh per query" and the 43.8% reduction. No standardized effect size applies.

"Our most energy-efficient configuration, DeepSeek-MoE (Ours), consumes 1.8 mWh per query, equivalent to 1.8 Wh per 1,000 queries, representing a 43.8% reduction relative to the Qwen-7B FP16 baseline (3.2 mWh/query)"

Discussion¶

Related Claims¶

• DeepSeek-MoE (Ours) achieves the strongest overall balance of accuracy among tested configurations, reaching 78.6% overall and 81.2% multi-hop reasoning accuracy on a 300-query OER evaluation set — related
• NF4 4-bit quantization reduces VRAM usage by 38.7% for Qwen-7B and 37.9% for DeepSeek-MoE while maintaining practical response times — related
• The prototype maintains total end-to-end latency of 3-5 seconds between query completion and RAG-generated response delivery — related
• Qwen 7B and Mistral Nemo 12B passed the classroom speed gate in round 2 and were retained as candidates pending faculty review — related