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
Run this task on Fronset — request an invitation
Models
Frontier on this task: Gemini 3.8 Flash at 8.21 / 10. Quality bar at 90%: 7.39.
point-estimate floor (CI low) · upper CI (less certain) · Bars sorted by blended cost; best-value model first.
| Model | Quality score | CI low | Cost / 1k runs | vs best value |
|---|---|---|---|---|
| Tencent Hy3 | 7.95 / 10 | 7.55 | $0.26 | best value |
| Qwen 3.7 Plus | 7.69 / 10 | 7.21 | $1.55 | 6x more expensive |
| Thinking Machines Inkling Small | 8.03 / 10 | 7.54 | $2.07 | 8x more expensive |
| Gemini 3.8 Flash | 8.21 / 10 | 7.74 | $2.28 | 8.8x more expensive |
| GLM-5.3 Flash | 8.18 / 10 | 7.70 | $2.29 | 8.8x more expensive |
| Meta Muse Spark 1.3 | 7.60 / 10 | 7.13 | $9.41 | 36x more expensive |
| Qwen 3.8 Max | 8.06 / 10 | 7.61 | $11.15 | 43x more expensive |
| Tencent Hy4 Preview | 7.48 / 10 | 6.99 | $17.47 | 67x more expensive |
| GLM-5.3 | 8.04 / 10 | 7.59 | $31.15 | 120x more expensive |
Cost breakdown
| Model | Quality | Confidence | Cost / 1k runs | Overpay | Mode |
|---|---|---|---|---|---|
| Tencent Hy3 ★ OpenRouter | 7.95 / 10 CI [7.55, 8.35] | MEDIUM | $0.26 | best value | batch |
| Qwen 3.7 Plus Alibaba Cloud (DashScope) | 7.69 / 10 CI [7.21, 8.16] | MEDIUM | $1.55 | 6x | batch |
| Thinking Machines Inkling Small OpenRouter | 8.03 / 10 CI [7.54, 8.51] | MEDIUM | $2.07 | 8x | batch |
| Gemini 3.8 Flash best Gemini | 8.21 / 10 CI [7.74, 8.69] | MEDIUM | $2.28 | 8.8x | batch |
| GLM-5.3 Flash Z.AI | 8.18 / 10 CI [7.70, 8.66] | MEDIUM | $2.29 | 8.8x | batch |
| Meta Muse Spark 1.3 OpenRouter | 7.60 / 10 CI [7.13, 8.07] | MEDIUM | $9.41 | 36x | batch |
| Qwen 3.8 Max Alibaba Cloud (DashScope) | 8.06 / 10 CI [7.61, 8.52] | MEDIUM | $11.15 | 43x | batch |
| Tencent Hy4 Preview OpenRouter | 7.48 / 10 CI [6.99, 7.96] | MEDIUM | $17.47 | 67x | batch |
| GLM-5.3 Z.AI | 8.04 / 10 CI [7.59, 8.49] | MEDIUM | $31.15 | 120x | 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: 1837 input tokens → 258 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 $/1M list rates — this is what running the task costs, not a per-token price.
Evaluation rubric
Judge primary and secondary section fit, ordered relevance, complete topic coverage, consistency, restraint in multi-assignment, and valid use of supplied sections.
Output schema
Every answer on this task is checked against this JSON Schema, whichever model wrote it. An answer that doesn't fit counts as a model failure, and the call is retried on another model.
{
"$defs": {
"SectionAssignment": {
"description": "Section assignment for a single merged topic.\n\nOrder is significant: index 0 is the PRIMARY section (the topic's single\nlanding-page home), index 1+ are SECONDARY sections (additional tag pages\nwhere the topic should also appear). Each topic appears once on the landing\npage under its primary section, and on every section's detail page.",
"properties": {
"section_ids": {
"description": "Ordered list of Ghost section IDs (1-3 sections). First ID = primary section (where the topic appears on the landing page). Remaining IDs = secondary sections, ordered by descending thematic relevance.",
"items": {
"type": "integer"
},
"title": "Section Ids",
"type": "array"
},
"topic_id": {
"description": "ID of the merged topic",
"title": "Topic Id",
"type": "integer"
}
},
"required": [
"topic_id"
],
"title": "SectionAssignment",
"type": "object"
}
},
"description": "Output from section assignment LLM call.",
"properties": {
"assignments": {
"description": "Section assignments for each topic in the batch",
"items": {
"$ref": "#/$defs/SectionAssignment"
},
"title": "Assignments",
"type": "array"
}
},
"title": "SectionAssignmentResult",
"type": "object"
}Prompt templates
The published template for this task.
default ·
LLMB_TOPIC_SECTION_ASSIGNMENT_SYSTEM +
LLMB_TOPIC_SECTION_ASSIGNMENT_USER
(100 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.