Best LLMs for Topic Section Assignment
Assigns each topic in a batch to one or more supplied sections, ordering assignments by relevance and preserving the topic as the unit of classification. Illustrative uses include assigning feedback themes to product areas, defects to software components, findings to audit sectio
Models
Frontier on this task: Gemini 3.5 Flash at 8.69 / 10. Quality bar at 90%: 7.82.
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 |
|---|---|---|---|---|
| GPT-5.6 Luna | 7.99 / 10 | 7.51 | $0.25 | best value |
| Tencent Hy3 | 8.12 / 10 | 7.72 | $0.28 | 1.1x more expensive |
| DeepSeek V4 Flash | 8.02 / 10 | 7.75 | $0.89 | 3.6x more expensive |
| GLM-5.3 Flash | 8.16 / 10 | 7.67 | $0.98 | 3.9x more expensive |
| GPT-5.6 Terra | 7.99 / 10 | 7.50 | $1.95 | 7.9x more expensive |
| Gemini 3.8 Flash | 8.16 / 10 | 7.70 | $2.30 | 9.3x more expensive |
| Thinking Machines Inkling Small | 8.11 / 10 | 7.63 | $2.46 | 9.9x more expensive |
| DeepSeek V4 Pro | 7.85 / 10 | 7.55 | $3.60 | 15x more expensive |
| Gemini 3.5 Flash | 8.69 / 10 | 8.38 | $4.63 | 19x more expensive |
| Qwen 3.8 Max | 8.05 / 10 | 7.59 | $11.95 | 48x more expensive |
| GLM-5.3 | 7.93 / 10 | 7.47 | $13.29 | 54x more expensive |
| Grok 4.6 | 8.08 / 10 | 7.58 | $14.58 | 59x more expensive |
| GPT-5.6 Sol | 8.26 / 10 | 7.78 | $21.53 | 87x more expensive |
| Meta Muse Spark 1.3 | 7.60 / 10 | 7.13 | $8.25 | 33x more expensive |
| Gemini 3.1 Flash Lite | 7.61 / 10 | 7.25 | $0.27 | 1.1x more expensive |
| GPT-5.4 Nano | 7.27 / 10 | 6.82 | $0.25 | 1x more expensive |
Cost breakdown
| Model | Quality | Confidence | Cost / 1k runs | Overpay | Mode |
|---|---|---|---|---|---|
| GPT-5.6 Luna ★ OpenAI | 7.99 / 10 CI [7.51, 8.48] | MEDIUM | $0.25 | best value | batch |
| Tencent Hy3 OpenRouter | 8.12 / 10 CI [7.72, 8.53] | MEDIUM | $0.28 | 1.1x | batch |
| DeepSeek V4 Flash DeepSeek | 8.02 / 10 CI [7.75, 8.30] | HIGH | $0.89 | 3.6x | batch |
| GLM-5.3 Flash Z.AI | 8.16 / 10 CI [7.67, 8.65] | MEDIUM | $0.98 | 3.9x | batch |
| GPT-5.6 Terra OpenAI | 7.99 / 10 CI [7.50, 8.48] | MEDIUM | $1.95 | 7.9x | batch |
| Gemini 3.8 Flash Gemini | 8.16 / 10 CI [7.70, 8.63] | MEDIUM | $2.30 | 9.3x | batch |
| Thinking Machines Inkling Small OpenRouter | 8.11 / 10 CI [7.63, 8.59] | MEDIUM | $2.46 | 9.9x | batch |
| DeepSeek V4 Pro DeepSeek | 7.85 / 10 CI [7.55, 8.14] | HIGH | $3.60 | 15x | batch |
| Gemini 3.5 Flash best Gemini | 8.69 / 10 CI [8.38, 9.00] | MEDIUM | $4.63 | 19x | batch |
| Qwen 3.8 Max Alibaba Cloud (DashScope) | 8.05 / 10 CI [7.59, 8.52] | MEDIUM | $11.95 | 48x | batch |
| GLM-5.3 Z.AI | 7.93 / 10 CI [7.47, 8.39] | MEDIUM | $13.29 | 54x | batch |
| Grok 4.6 xAI | 8.08 / 10 CI [7.58, 8.58] | MEDIUM | $14.58 | 59x | batch |
| GPT-5.6 Sol OpenAI | 8.26 / 10 CI [7.78, 8.75] | MEDIUM | $21.53 | 87x | 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: 1849 input tokens → 1737 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 primary and secondary section fit, ordered relevance, complete topic coverage, consistency, restraint in multi-assignment, and valid use of supplied sections.
Prompt templates
This is a pooled capability — 2 prompt families share it. The pair shown first is the most frequently used in production.
TOPIC_CLUSTERING_ASSIGN_SECTIONS_SYSTEM +
TOPIC_CLUSTERING_ASSIGN_SECTIONS_USER
(627 calls in window)
System prompt
You are an expert content analyst. Your task is to assign publication sections to a batch of merged topics.
## Section Assignment Guidelines:
1. Each topic should appear in 1-3 sections (not more)
2. Match topic themes to section purposes
3. Consider the section's audience and goals
4. Don't force-fit topics into sections
5. Be consistent with section assignments across batches
## Ordering (important):
The `section_ids` list is ORDERED by descending relevance.
- Index 0 = PRIMARY section: the single most thematically central section.
This is where the topic will appear on the client's landing page.
- Index 1+ = SECONDARY sections: additional sections where the topic also
belongs and should appear on the section's detail page.
Pick the primary as the section a reader would most expect to find this
topic under. Use secondary slots only when the topic genuinely spans
multiple sections.
## Output Requirements:
- Every topic in the batch must have a section assignment
- Assign 1-3 section_ids per topic based on relevance, ordered as described above
- Use the actual section IDs from the available sections list
## Required Output Format
Your response MUST be a single, valid JSON object conforming to this schema:
```json
{schema_json_string}
```User prompt
Please assign publication sections to this batch of merged topics.
## Topics to Assign (Batch {batch_number} of {total_batches}):
{batch_topics_json}
## Available Sections:
{sections_json}
## Required JSON Schema:
The required JSON output schema is provided in the system prompt.
## Important:
- Assign 1-3 sections per topic
- Every topic in this batch must have section assignments
- Use actual section IDs from the available sections list
- Order matters: list the most thematically central section FIRST (it becomes
the primary, where the topic appears on the landing page); list any
additional sections after it in descending relevance.
LLMB_TOPIC_SECTION_ASSIGNMENT_SYSTEM +
LLMB_TOPIC_SECTION_ASSIGNMENT_USER
(37 calls in window)
System prompt
Compare each topic’s title and description with section definitions. Select the strongest supported section first and additional sections only when they reflect material secondary scope. Do not infer fit from the section’s popularity or access level, create section IDs, or leave an input topic unaccounted for. Treat empty optional values as absent and return only the requested result. Your response must conform exactly to this output schema: {schema_json_string}.
User prompt
Inputs — batch_number: {batch_number}; total_batches: {total_batches}; batch_topics_json: {batch_topics_json}; sections_json: {sections_json}. Use only these inputs to complete the task defined by the system prompt.