{
  "abstract": "Deploying large language models (LLMs) in clinical settings is limited by security, reliability, latency, and accessibility concerns that favor smaller, on-device or on-premise models. However, these smaller models may struggle to meet accuracy requirements. While fine-tuning and retrieval-augmented generation (RAG) can improve domain-specific accuracy, these methods require additional labeled data, technical skill, and infrastructure. In contrast, test-time scaling—allocating extra token-budget during inference—offers a training-free alternative to increasing accuracy. However, the trade-offs between these strategies and their interaction with model size remain poorly understood for medical reasoning. To address this gap, we compare three approaches—test-time scaling, fine-tuning, and context grounding—using the Gemma and MedGemma family of LLMs (Gemma-3 1B, Gemma-3 4B, Gemma-3 27B, MedGemma-4B, and MedGemma 27B) and evaluate these systems across common biomedical question-answering (QA) datasets and a set of recently released medical exam questions with the performance of practicing clinicians available for comparison. We test baseline prompts (direct answer, Chain-of-Thought, and self-consistency) while introducing a new prompting method we call “prompt-chaining for continuous reflection” (PCCR) that forces inference time minimum token-generation budgets. We assess accuracy and tokens-generated, allowing us to investigate the accuracy–efficiency trade-offs across prompting, context-grounding, fine-tuning, and model scales. We discover equivalency point configurations where a smaller model’s accuracy falls within one 95% confidence interval of a larger model’s (typically within 1–4 percentage points) reached through increased reasoning budgets, context-grounding, or fine-tuning. Specific effects are apparent and statistically supported: the benefit of medical fine-tuning grew from +4.6 to +15.7 percentage points (non-overlapping 95% CIs) when paired with self-consistency, and enforced extended reasoning raised MedGemma 27B accuracy from 58.1% to 80.1% (p < 10-5) in the absence of context. We also identify an “overthinking” inflection: when high-quality context is available, extended reasoning beyond roughly 128–256 tokens degrades accuracy by 7–13 percentage points. Using these empirical results, we formulate a general framework with equations to balance cost-benefit trade-offs when engineering LLM-based systems for medical reasoning and QA. We recommend generalizable configurations, designs, and patterns to achieve accuracy and efficiency objectives for example use-cases relevant to healthcare organizations. Author summary: When doctors and hospitals want to use artificial intelligence for medical tasks, they face difficult choices. The most capable AI systems are expensive to run and require sending sensitive patient and hospital data to external servers. Smaller systems that can run locally are more practical but may be less accurate. Our research asked: with what configurations can we make smaller AI systems perform as well as larger ones for medical reasoning? We tested five AI models of different sizes, including both generalist and medically-specialized models, on thousands of medical questions using various prompting strategies, including a new and very simple method we developed that encourages the AI to reason more extensively. We discovered several surprising findings which we describe in detail. Most importantly, we found that smaller, cheaper models can match larger ones when given the right combination of prompting strategy, specialized training, and supporting information. We translated these findings into practical guidelines that can help choose AI configurations that balance accuracy, speed, and cost. Our framework could help make medical AI more accessible to institutions with limited computational resources while achieving the highest possible accuracy.",
  "abstract_shown": true,
  "affiliations": [
    "SKA Labs, Atlanta, Georgia, United States of America",
    "Advanced Product Development, Intuitive Surgical, Inc., Atlanta, Georgia, United States of America"
  ],
  "article_type": "Research Article",
  "arxiv": null,
  "author_total": 2,
  "authors": [
    {
      "affs": [
        0
      ],
      "name": "Kiran Bhattacharyya"
    },
    {
      "affs": [
        1
      ],
      "name": "Sreeram Kamabattula"
    }
  ],
  "bibtex": "https://research.draconitic.com/paper/51jr1686m5765.bib",
  "cited_by": 0,
  "doi": "10.1371/journal.pdig.0001182",
  "fulltext_url": "https://journals.plos.org/digitalhealth/article?id=10.1371/journal.pdig.0001182",
  "id": "51jr1686m5765",
  "issns": [
    "2767-3170"
  ],
  "landing_url": "https://journals.plos.org/digitalhealth/article?id=10.1371/journal.pdig.0001182",
  "language": null,
  "licence": "CC-BY-4.0",
  "licence_label": "CC BY 4.0",
  "licence_source": "journal",
  "licence_url": "https://creativecommons.org/licenses/by/4.0/",
  "metadata_licence": null,
  "native_subjects": [
    [
      "openalex",
      "2718|Artificial Intelligence in Healthcare and Education"
    ],
    [
      "plos",
      "/Biology and life sciences/Neuroscience/Cognitive science/Cognitive psychology/Language"
    ],
    [
      "plos",
      "/Biology and life sciences/Neuroscience/Cognitive science/Cognitive psychology/Reasoning"
    ],
    [
      "plos",
      "/Biology and life sciences/Psychology/Cognitive psychology/Language"
    ],
    [
      "plos",
      "/Biology and life sciences/Psychology/Cognitive psychology/Reasoning"
    ],
    [
      "plos",
      "/Biology and life sciences/Psychology/Psychometrics"
    ],
    [
      "plos",
      "/Computer and information sciences/Artificial intelligence"
    ],
    [
      "plos",
      "/Medicine and health sciences/Health care/Health care providers/Physicians"
    ],
    [
      "plos",
      "/Medicine and health sciences/Radiology and imaging"
    ],
    [
      "plos",
      "/People and places/Population groupings/Professions/Medical personnel/Physicians"
    ],
    [
      "plos",
      "/Physical sciences/Physics/Classical mechanics/Reflection"
    ],
    [
      "plos",
      "/Social sciences/Economics/Economic models/Cost models"
    ],
    [
      "plos",
      "/Social sciences/Psychology/Cognitive psychology/Language"
    ],
    [
      "plos",
      "/Social sciences/Psychology/Cognitive psychology/Reasoning"
    ],
    [
      "plos",
      "/Social sciences/Psychology/Psychometrics"
    ]
  ],
  "openalex": "W7213545303",
  "pdf_url": "https://journals.plos.org/digitalhealth/article/file?id=10.1371/journal.pdig.0001182&type=printable",
  "pmcid": null,
  "pmid": null,
  "published": "2026-09-18",
  "publisher": "Public Library of Science",
  "ris": "https://research.draconitic.com/paper/51jr1686m5765.ris",
  "sources": [
    {
      "first_seen": "2026-10-30T12:17:43Z",
      "native_id": "10.1371/journal.pdig.0001182",
      "source": "plos",
      "url": "https://journals.plos.org/digitalhealth/article?id=10.1371/journal.pdig.0001182"
    }
  ],
  "subjects": [
    {
      "field": "biology",
      "origin": "plos:neuroscience",
      "subfield": "neuroscience"
    },
    {
      "field": "medicine",
      "origin": "plos:radiology and imaging",
      "subfield": "imaging"
    }
  ],
  "title": "The economics of accuracy for medical reasoning with large language models",
  "title_text": "The economics of accuracy for medical reasoning with large language models",
  "type": "peer-reviewed",
  "updated_at": "2026-10-30T12:17:49Z",
  "url": "https://research.draconitic.com/paper/51jr1686m5765",
  "venue": "PLOS Digital Health",
  "versions": []
}