Cost mode:

Category: Content Summarization & Synthesis · Rail: absolute · Typical I/O: 34028→7544 tokens

Models

Frontier on this task: GLM-5.3 at 8.71 / 10. Quality bar at 90%: 7.84.

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 clears the bar but whose CI low does not.

ModelQuality scoreCI lowCost / 1k runsvs best value
GPT-5.4 Nano7.84 / 107.59$15.71best value
Qwen 3.7 Plus8.18 / 107.98$22.351.4x more expensive
Gemini 3.8 Flash8.15 / 107.84$23.451.5x more expensive
Claude Sonnet 58.21 / 108.10$45.492.9x more expensive
GPT-5.6 Terra7.95 / 107.71$49.623.2x more expensive
Claude Haiku 4.58.01 / 107.70$51.373.3x more expensive
Grok 4.68.37 / 108.10$116.137.4x more expensive
DeepSeek V4 Pro8.28 / 107.87$122.747.8x more expensive
GLM-5.38.71 / 108.51$122.847.8x more expensive
Moonshot Kimi K37.97 / 107.74$185.5412x more expensive
Thinking Machines Inkling Small7.69 / 107.35$30.301.9x more expensive
Meta Muse Spark 1.37.72 / 107.34$65.274.2x more expensive
Gemini 3.5 Flash7.74 / 107.40$34.202.2x more expensive
Gemini 3.5 Flash Lite6.89 / 106.42$7.5552% cheaper
GPT-5.6 Luna7.84 / 107.56$4.7170% cheaper
MiniMax M37.61 / 107.46$26.931.7x more expensive
Thinking Machines Inkling7.67 / 107.33$50.313.2x more expensive

Cost breakdown

ModelQualityConfidenceCost / 1k runsOverpayMode
GPT-5.4 Nano OpenAI7.84 / 10 CI [7.59, 8.09]HIGH$15.71best valuebatch
Qwen 3.7 Plus Alibaba Cloud (DashScope)8.18 / 10 CI [7.98, 8.37]RANKED$22.351.4xbatch
Gemini 3.8 Flash Gemini8.15 / 10 CI [7.84, 8.47]MEDIUM$23.451.5xbatch
Claude Sonnet 5 Anthropic8.21 / 10 CI [8.10, 8.32]RANKED$45.492.9xbatch
GPT-5.6 Terra OpenAI7.95 / 10 CI [7.71, 8.18]HIGH$49.623.2xbatch
Claude Haiku 4.5 Anthropic8.01 / 10 CI [7.70, 8.31]MEDIUM$51.373.3xbatch
Grok 4.6 xAI8.37 / 10 CI [8.10, 8.63]HIGH$116.137.4xbatch
DeepSeek V4 Pro DeepSeek8.28 / 10 CI [7.87, 8.68]MEDIUM$122.747.8xbatch
GLM-5.3 best Z.AI8.71 / 10 CI [8.51, 8.91]HIGH$122.847.8xbatch
Moonshot Kimi K3 Moonshot AI7.97 / 10 CI [7.74, 8.21]HIGH$185.5412xbatch

Overpay shows how much more you pay than the best-value model that clears the quality bar (marked ★) — the best-value good-enough option. "16x" means you overpay 16× — 16× that reference for no quality benefit above the bar. Typical call shape for this task: 34028 input tokens → 7544 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 coverage of decision-relevant content, factual fidelity, prioritization, preservation of caveats, usefulness to the stated audience, coherence, and compliance with the semantic length target. Exact word count is deterministic.

Prompt templates

The system + user template pair used for this task.

EXECUTIVE_SUMMARY_SYSTEM_PROMPT + EXECUTIVE_SUMMARY_USER_PROMPT (493 calls in window)

System prompt

You are an expert financial analyst specializing in creating executive summaries for research reports.

Your task is to write a concise, compelling executive summary that:
- Highlights the most critical insights from the full report
- Identifies key trends, risks, and opportunities
- Provides actionable recommendations for decision-makers
- Uses professional, analytical tone
- Matches the target length specified in the user prompt (typically 5-10% of the full report)

Focus on the big picture. Avoid detailed data points - those belong in the chapters below.

The length target will be provided as approximate word count. Adjust the number of paragraphs as needed to match the target length while maintaining coherent structure.

## Required Output Format
Your response MUST be a single, valid JSON object conforming to this schema:
```json
{schema_json_string}
```

User prompt

Subject: {subject_name} ({subject_code})

Report Requirements:
{report_requirements}

Consolidated Chapters (approximately {chapters_word_count} words):
{consolidated_chapters}

Task: Write an executive summary that synthesizes the key insights from all chapters above.

Target Length: Approximately {target_word_count} words ({target_percentage}% of the full report).

The required JSON output schema is provided in the system prompt.