A Picture is Worth a Thousand Tokens: How Vision Language Models Cut AI Energy Costs While Improving Accuracy
On the map: Multimodal LLMs / vision-language models → Visual token pruning and efficiency
Abstract (arXiv)
LLM inference accounts for over 90% of AI operational energy, scaling directly with input token count---a critical inefficiency for telecom network analytics and numerical time-series data analysis (NTSDA), where raw multivariate KPI windows from 4G/5G cell sites expand into thousands of floating-point tokens. Vision-Language Models (VLMs) eliminate this mismatch by encoding time-series as 2D plots, achieving 3.6-10.4x input token reduction across Llama-3.2-90B, Qwen2.5-VL-72B, and Pixtral-12B architectures. This translates to 1.8-2.5x measured inference energy reduction, saving approximately 7.2 MJ/day at telecom edge deployments and CloudRAN that monitor 200 cells per 15-minute interval. Critically, efficiency gains do not sacrifice accuracy: a fine-tuned Llama-3.2-90B-Vision VLM achieves 220.7% higher precision than its text-only counterpart and outperforms LSTM and ARIMA baselines by over 144% on telecom anomaly detection. On public benchmarks, Pixtral-12B achieves a 20.6x improvement in J/F1 score at mean F1 = 0.82. At 24 KPIs, text representations exceed the 128K context window of most production LLMs, rendering text-only processing infeasible without truncation, while visual representations remain within standard limits. These results establish VLMs as an energy-efficient and accuracy-superior modality for numerical time-series workloads, providing empirical grounding for AI inference systems that treat energy consumption as a first-class engineering constraint.
Key facts. No citing papers in the map yet (data through 14 Oct 2026). Placed in Visual token pruning and efficiency, one of 16 directions in Multimodal LLMs / vision-language models. Builds on 15 of its 23 references that are in the map. First posted on arXiv 7 Aug 2026.
Builds on
15 of 23 references.
Closest ancestors · 1
Same direction, newest first
Other references · 14
- 2024The Llama 3 Herd of Modelscited by 14,715
- 2024Qwen2-VL: Enhancing Vision-Language Model's Perception of the World at Any Resolutioncited by 4,234
- 2024Pixtral 12Bcited by 179
- 2024Tokenization counts: the impact of tokenization on arithmetic in frontier LLMscited by 115
- 2022Zeus: Understanding and Optimizing GPU Energy Consumption of DNN Trainingcited by 79
- 2025How Hungry is AI? Benchmarking Energy, Water, and Carbon Footprint of LLM Inferencecited by 76
- 2024A Picture is Worth A Thousand Numbers: Enabling LLMs Reason about Time Series via Visualizationcited by 33
- 2025The ML.ENERGY Benchmark: Toward Automated Inference Energy Measurement and Optimizationcited by 33
- 2025TokenPowerBench: Benchmarking the Power Consumption of LLM Inferencecited by 22
- 2025From Prompts to Power: Measuring the Energy Footprint of LLM Inferencecited by 12
- 2025Harnessing Vision-Language Models for Time Series Anomaly Detectioncited by 11
- 2025How Different Tokenization Algorithms Impact LLMs and Transformer Models for Binary Code Analysiscited by 3
Built on by
Not cited yet.
Similar, not cited
Close in wording, no citation either way
- 2026Position: LLM Inference Should Be Evaluated as Energy-to-Token Production
- 2025Sustainable LLM Inference for Edge AI: Evaluating Quantized LLMs for Energy Efficiency, Output Accuracy, and Inference Latency
- 2026Beyond Test-Time Compute Strategies: Advocating Energy-per-Token in LLM Inference
- 2026Characterization of Request and Token Energy Costs for LLM Inference Workloads on GPU Platforms
- 2024Inference Optimal VLMs Need Fewer Visual Tokens and More Parameters
- 2026Seeing is Free, Speaking is Not: Uncovering the True Energy Bottleneck in Edge VLM Inference
- 2025Characterizing LLM Inference Energy-Performance Tradeoffs across Workloads and GPU Scaling
- 2026Understanding Efficiency: Quantization, Batching, and Serving Strategies in LLM Energy Use