Fronset Benchmark

Best LLMs for Structured Output Extraction

Extracts structured data from text into a specified JSON schema. Pure shape-conformance — no enrichment, no rephrasing, no summarisation. Used when a downstream consumer needs schema-clean data from prose.

What the steps mean
Cost mode:

Category: Structured Data & Fact Extraction · Typical I/O: 3211→12079 tokens

Runs on model group: All models. Inputs may reach any model in the group, not only the ranked models below.

Run this task on Fronset — request an invitation

Models

Frontier on this task: Claude Opus 5.5 at 9.93 / 10. Score needed at the Balanced step: 8.94.

point-estimate floor (CI low) · upper CI (less certain) · Bars sorted by blended cost; best-value model first. Greyed rows are MEDIUM+ models whose point estimate does not clear the bar.

Cost breakdown

ModelQualityConfidenceCost / 1k runsOverpayMode
Xiaomi MiMo V2.6 Flash ★ Xiaomi MiMo9.76 / 10 CI [9.54, 9.98]HIGH$0.74best valuebatch
GPT-6 Luna OpenAI9.11 / 10 CI [8.84, 9.38]HIGH$1.912.6xbatch
Xiaomi MiMo V2.6 Pro Xiaomi MiMo9.64 / 10 CI [9.38, 9.89]HIGH$2.203xbatch
DeepSeek V4.1 Flash DeepSeek9.25 / 10 CI [8.92, 9.58]MEDIUM$3.034.1xbatch
MiniMax M3 OpenRouter9.12 / 10 CI [8.97, 9.28]RANKED$4.526.1xbatch
GPT-6 Sol OpenAI9.20 / 10 CI [8.88, 9.53]MEDIUM$23.4232xbatch
Qwen 3.7 Plus Alibaba Cloud (DashScope)9.56 / 10 CI [9.40, 9.72]RANKED$26.9336xbatch
Claude Sonnet 5.5 Anthropic9.88 / 10 CI [9.71, 10.00]RANKED$33.9246xbatch
Claude Opus 5.5 best Anthropic9.93 / 10 CI [9.77, 10.00]RANKED$68.0792xbatch
Moonshot Kimi K3 Moonshot AI9.59 / 10 CI [9.29, 9.89]HIGH$82.63112xbatch
Qwen 3.8 Max Alibaba Cloud (DashScope)9.32 / 10 CI [8.93, 9.71]MEDIUM$139.28188xbatch

Overpay shows how much more you pay than the best-value model that meets the quality step (marked ★) — the best-value good-enough option. "16x" means you overpay 16× — 16× that reference for no quality benefit the step asks for. Typical call shape for this task: 3211 input tokens → 12079 output tokens, EMA-tracked from production traffic. Cost is the observed, all-in $ per 1,000 task runs: each model's own measured usage on this task — output verbosity, thinking/reasoning tokens, cache reads and writes, and the spend on its billed failures — priced at current list rates and adjusted by the billing overhead we actually reconcile against provider invoices. Models that answer tersely cost what they actually cost; models that think at length pay for it. Not comparable to providers' advertised $/1M list rates — this is what running the task costs, not a per-token price.

Evaluation rubric

Judge field-level source support, completeness for present values, correct missing-value handling, semantic and quantitative fidelity, resistance to embedded instructions, and absence of enrichment. JSON/schema validity is deterministic.

Output schema

Every answer on this task is checked against this JSON Schema, whichever model wrote it. An answer that doesn't fit counts as a model failure, and the call is retried on another model.

{
  "properties": {
    "report_markdown": {
      "description": "The full synthesized report formatted in Markdown.",
      "title": "Report Markdown",
      "type": "string"
    }
  },
  "required": [
    "report_markdown"
  ],
  "title": "SynthesizedOutputMarkdownOnly",
  "type": "object"
}

Prompt templates

The published template for this task.

default · STRUCTURED_OUTPUT_EXTRACTION_SYSTEM_PROMPT + STRUCTURED_OUTPUT_EXTRACTION_USER_PROMPT (2695 calls in window)

System prompt

You are a JSON extraction assistant. Extract structured data from the provided text that matches the given JSON schema. Output ONLY the structured data matching the schema.

User prompt

Extract the structured data from the following text that matches this JSON schema:

## JSON Schema:
```json
{schema_json}
```

## Text to extract from:
{raw_text}