{
  "$schema": "https://themachinepress.com/schemas/story-v1.schema.json",
  "schema_version": "1.0.0",
  "document_type": "machine_press_story",
  "story": {
    "story_id": "mp-2026-09-09-004",
    "source_story_id": "tmp-story-darebench-agent-workload-matrix",
    "edition_id": "mp-2026-09-09-morning-0062",
    "edition_url": "https://themachinepress.com/edition/2026-09-09",
    "position": 4,
    "story_type": "dispatch",
    "section": "frontier-models",
    "editorial_classification": "editorial",
    "headline": "No Agent Model Won Every Kind of Work",
    "slug": "no-agent-model-won-every-kind-of-work",
    "dek": "DAREBench placed 233 tasks into six workload groups and compared 35 commercial and open models over 7,587 runs.",
    "summary": "DAREBench placed 233 tasks into six workload groups and compared 35 commercial and open models over 7,587 runs.",
    "body_text": "DAREBench adapts tasks from 22 source benchmarks into a two-by-three matrix defined by input modality and execution form. All tasks run in a shared environment with contract-based scoring and evidence audits. Across 23 commercial API models and 12 locally deployed open-weight models, no system dominated all groups. Text and multimodal workloads produced different accuracy-cost trade-offs, while local models were competitive in several groups but trailed frontier commercial systems overall. The benchmark argues that deployment choices should follow workload profiles rather than a single aggregate score; its conclusions remain tied to the selected tasks, environment and reference cost assumptions.",
    "why_it_matters": "DAREBench placed 233 tasks into six workload groups and compared 35 commercial and open models over 7,587 runs.",
    "limitations": [
      "The benchmark argues that deployment choices should follow workload profiles rather than a single aggregate score; its conclusions remain tied to the selected tasks, environment and reference cost assumptions."
    ],
    "importance": 8,
    "canonical_url": "https://themachinepress.com/story/mp-2026-09-09-004/no-agent-model-won-every-kind-of-work",
    "json_url": "https://themachinepress.com/story/mp-2026-09-09-004.json",
    "first_published_at": "2026-09-09T09:00:00.000-04:00",
    "modified_at": "2026-09-09T09:00:00.000-04:00",
    "content_status": "new",
    "is_carryover": false,
    "carryover_reason": null,
    "key_claims": [
      {
        "claim_id": "claim-mp-2026-09-09-004-001",
        "text": "DAREBench placed 233 tasks into six workload groups and compared 35 commercial and open models over 7,587 runs.",
        "source_ids": [
          "source-2026-09-09-004"
        ],
        "qualification": "The benchmark argues that deployment choices should follow workload profiles rather than a single aggregate score; its conclusions remain tied to the selected tasks, environment and reference cost assumptions."
      }
    ],
    "source_ids": [
      "source-2026-09-09-004"
    ],
    "tags": [
      "agent benchmarks",
      "model selection",
      "deployment"
    ],
    "image_url": null,
    "corrections": []
  },
  "sources": [
    {
      "source_id": "source-2026-09-09-004",
      "title": "arXiv preprint 2609.06059",
      "publisher": "arXiv",
      "url": "https://arxiv.org/abs/2609.06059",
      "canonical_url": "https://arxiv.org/abs/2609.06059",
      "source_type": "primary_research",
      "is_primary_source": true,
      "published_at": "2026-09-05T08:36:59.000-04:00",
      "accessed_at": "2026-09-09T08:30:00.000-04:00",
      "supports_claim_ids": [
        "claim-mp-2026-09-09-004-001"
      ]
    }
  ],
  "corrections": [],
  "publisher": {
    "name": "The Machine Press",
    "url": "https://themachinepress.com",
    "description": "A daily newspaper for the age of artificial intelligence."
  },
  "cite_this_report": {
    "title": "No Agent Model Won Every Kind of Work",
    "publisher": "The Machine Press",
    "published_at": "2026-09-09T09:00:00.000-04:00",
    "canonical_url": "https://themachinepress.com/story/mp-2026-09-09-004/no-agent-model-won-every-kind-of-work"
  }
}