Best LLMs for SEC S-1 Chunk Analysis
Per-section analysis of an S-1 / S-1/A registration statement for a long-term investor: business model, financial metrics, risk factors, strategic direction, market opportunity. Grounded only in the section's text — no outside knowledge.
Models
Frontier on this task: Claude Opus 4.8 at 9.21 / 10. Quality bar at 90%: 8.29.
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.
| Model | Quality score | CI low | Cost / 1k runs | vs best value |
|---|---|---|---|---|
| Qwen 3.5 Flash | 8.48 / 10 | 8.28 | $3.11 | best value |
| GPT-5.4 Nano | 8.45 / 10 | 8.30 | $4.81 | 1.5x more expensive |
| MiniMax M3 | 9.06 / 10 | 9.01 | $5.70 | 1.8x more expensive |
| GPT-5.6 Luna | 8.86 / 10 | 8.61 | $7.68 | 2.5x more expensive |
| Qwen 3.7 Plus | 8.59 / 10 | 8.43 | $9.53 | 3.1x more expensive |
| Qwen 3.6 Flash | 8.45 / 10 | 8.33 | $11.33 | 3.6x more expensive |
| Qwen 3.6 Plus | 8.70 / 10 | 8.54 | $15.53 | 5x more expensive |
| Gemini 3.5 Flash | 8.97 / 10 | 8.87 | $18.79 | 6x more expensive |
| GPT-5.6 Terra | 8.96 / 10 | 8.74 | $20.12 | 6.5x more expensive |
| Grok 4.5 | 8.99 / 10 | 8.93 | $29.39 | 9.5x more expensive |
| Meta Muse Spark 1.1 | 9.20 / 10 | 8.97 | $31.06 | 10x more expensive |
| Claude Sonnet 5 | 8.89 / 10 | 8.78 | $35.28 | 11x more expensive |
| Kimi K2.6 | 8.83 / 10 | 8.69 | $36.76 | 12x more expensive |
| GPT-5.6 Sol | 9.17 / 10 | 8.94 | $42.50 | 14x more expensive |
| Claude Opus 4.8 | 9.21 / 10 | 8.93 | $53.86 | 17x more expensive |
| GPT-5.5 | 9.04 / 10 | 8.94 | $87.94 | 28x more expensive |
| GPT-5.4 Mini | 8.03 / 10 | 7.76 | $5.69 | 1.8x more expensive |
| Gemini 3.1 Pro Preview | 7.64 / 10 | 7.42 | $12.77 | 4.1x more expensive |
| DeepSeek V4 Pro | 7.93 / 10 | 7.53 | $8.97 | 2.9x more expensive |
| Claude Sonnet 4.6 | 7.82 / 10 | 7.45 | $67.08 | 22x more expensive |
| Claude Haiku 4.5 | 8.08 / 10 | 7.79 | $18.54 | 6x more expensive |
| Gemini 3.1 Flash Lite | 7.47 / 10 | 7.23 | $2.28 | 27% cheaper |
| DeepSeek V4 Flash | 8.24 / 10 | 7.95 | $1.35 | 56% cheaper |
| Tencent Hy3 | 7.85 / 10 | 7.50 | $4.02 | 1.3x more expensive |
Cost breakdown
| Model | Quality | Confidence | Cost / 1k runs | Overpay | Mode |
|---|---|---|---|---|---|
| Qwen 3.5 Flash ★ Alibaba Cloud (DashScope) | 8.48 / 10 CI [8.28, 8.68] | RANKED | $3.11 | best value | batch |
| GPT-5.4 Nano OpenAI | 8.45 / 10 CI [8.30, 8.60] | RANKED | $4.81 | 1.5x | batch |
| MiniMax M3 MiniMax | 9.06 / 10 CI [9.01, 9.11] | RANKED | $5.70 | 1.8x | batch |
| GPT-5.6 Luna OpenAI | 8.86 / 10 CI [8.61, 9.11] | HIGH | $7.68 | 2.5x | batch |
| Qwen 3.7 Plus Alibaba Cloud (DashScope) | 8.59 / 10 CI [8.43, 8.76] | RANKED | $9.53 | 3.1x | batch |
| Qwen 3.6 Flash Alibaba Cloud (DashScope) | 8.45 / 10 CI [8.33, 8.56] | RANKED | $11.33 | 3.6x | batch |
| Qwen 3.6 Plus Alibaba Cloud (DashScope) | 8.70 / 10 CI [8.54, 8.86] | RANKED | $15.53 | 5x | batch |
| Gemini 3.5 Flash Gemini | 8.97 / 10 CI [8.87, 9.07] | RANKED | $18.79 | 6x | batch |
| GPT-5.6 Terra OpenAI | 8.96 / 10 CI [8.74, 9.17] | HIGH | $20.12 | 6.5x | batch |
| Grok 4.5 xAI | 8.99 / 10 CI [8.93, 9.05] | RANKED | $29.39 | 9.5x | batch |
| Meta Muse Spark 1.1 Meta | 9.20 / 10 CI [8.97, 9.43] | HIGH | $31.06 | 10x | batch |
| Claude Sonnet 5 Anthropic | 8.89 / 10 CI [8.78, 8.99] | RANKED | $35.28 | 11x | batch |
| Kimi K2.6 Moonshot AI | 8.83 / 10 CI [8.69, 8.98] | RANKED | $36.76 | 12x | batch |
| GPT-5.6 Sol OpenAI | 9.17 / 10 CI [8.94, 9.39] | HIGH | $42.50 | 14x | batch |
| Claude Opus 4.8 best Anthropic | 9.21 / 10 CI [8.93, 9.50] | HIGH | $53.86 | 17x | batch |
| GPT-5.5 OpenAI | 9.04 / 10 CI [8.94, 9.14] | RANKED | $87.94 | 28x | batch |
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: 13100 input tokens → 1754 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.
Prompt templates
This is a pooled capability — 2 prompt families share it. The pair shown first is the most frequently used in production.
SEC_S1_CHUNK_ANALYSIS_SYSTEM_PROMPT +
SEC_S1_CHUNK_USER_PROMPT
(2343 calls in window)
System prompt
You are a senior investment analyst at a long-term focused investment firm. You specialize in analyzing SEC filings, particularly S-1 and S-1/A registration statements for companies going public.
You will be provided with a specific section from an S-1 filing. Your job is to extract the most investment-relevant information from this section and analyze its implications for long-term investors.
Focus on:
- Business model insights and competitive positioning
- Financial performance and metrics
- Risk factors and potential concerns
- Strategic direction and management quality
- Market opportunity and growth prospects
Provide your analysis as a structured JSON object using only information found directly in the section. Do not speculate or add external knowledge.
## Required Output Format
Your response MUST be a single, valid JSON object conforming to this schema:
```json
{schema_json_string}
```User prompt
Analyze the following section from an S-1 filing: **{section_title}**
**Section Content:**
```
{section_content}
```
**Instructions:**
1. Extract the most important investment-relevant information from this section
2. Focus on insights that would help a long-term investor evaluate this company
3. Identify any business model insights, financial information, risks, or competitive factors
4. Your analysis should be concise but comprehensive
5. Use only information directly stated in the section content
**JSON Output Format:**
The required JSON output schema is provided in the system prompt.JSON_REPAIR_SYSTEM +
JSON_REPAIR_USER
(5 calls in window)
System prompt
You are a JSON repair tool. The user gives you malformed or partial model output and a JSON Schema. Return ONLY a single valid JSON object that satisfies the schema, salvaging as much real content from the input as possible. Do not invent data for fields the input doesn't support — use the schema's allowed empty/null values. Output the JSON object only: no prose, no markdown, no code fences.
User prompt
JSON Schema:
{schema_json}
Malformed output to repair:
{raw_text}
Return only the corrected JSON object.