Best LLMs for Topic Sequence Optimization
Orders supplied topics for a coherent long-form presentation using logical dependencies, increasing complexity, narrative flow, and reader value. Illustrative uses include ordering an implementation plan, customer guide, software course, board narrative, research report, document
Run this task on Fronset — request an invitation
Models
Frontier on this task: GLM-5.3 at 8.88 / 10. Quality bar at 90%: 7.99.
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.
| Model | Quality score | CI low | Cost / 1k runs | vs best value |
|---|---|---|---|---|
| Xiaomi MiMo V2.6 Flash | 8.17 / 10 | 7.91 | $0.24 | best value |
| Tencent Hy3 | 8.24 / 10 | 7.96 | $0.77 | 3.2x more expensive |
| Xiaomi MiMo V2.6 Pro | 8.51 / 10 | 8.25 | $1.10 | 4.5x more expensive |
| NVIDIA Nemotron-3 Ultra 550B | 8.09 / 10 | 7.72 | $5.38 | 22x more expensive |
| GLM-5.3 Flash | 8.59 / 10 | 8.38 | $15.05 | 62x more expensive |
| Meta Muse Spark 1.3 | 8.06 / 10 | 7.84 | $18.07 | 74x more expensive |
| Thinking Machines Inkling | 8.01 / 10 | 7.71 | $33.17 | 136x more expensive |
| Grok 4.7 | 8.28 / 10 | 7.96 | $52.92 | 217x more expensive |
| Moonshot Kimi K3 | 8.48 / 10 | 8.16 | $71.69 | 295x more expensive |
| Qwen 3.8 Max | 8.07 / 10 | 7.77 | $107.70 | 443x more expensive |
| GLM-5.3 | 8.88 / 10 | 8.66 | $175.06 | 719x more expensive |
| Claude Opus 5.5 | 7.81 / 10 | 7.46 | $24.14 | 99x more expensive |
| Tencent Hy4 Preview | 7.74 / 10 | 7.34 | $46.42 | 191x more expensive |
| Gemini 3.8 Flash | 7.99 / 10 | 7.67 | $6.01 | 25x more expensive |
| GPT-6 Sol | 7.18 / 10 | 6.88 | $6.95 | 29x more expensive |
| Claude Haiku 5.5 | 7.20 / 10 | 6.86 | $0.74 | 3x more expensive |
| Qwen 3.8 Flash | 7.60 / 10 | 7.19 | $12.35 | 51x more expensive |
| Gemini 3.5 Flash Lite | 7.07 / 10 | 6.79 | $1.79 | 7.4x more expensive |
| NVIDIA Nemotron-3 Nano 30B-A3B | 6.71 / 10 | 6.24 | $3.17 | 13x more expensive |
| MiniMax M3 | 7.81 / 10 | 7.39 | $0.96 | 3.9x more expensive |
| NVIDIA Nemotron-3 Super 120B | 7.30 / 10 | 6.90 | $2.06 | 8.5x more expensive |
| Qwen 3.7 Plus | 7.82 / 10 | 7.43 | $5.39 | 22x more expensive |
| Thinking Machines Inkling Small | 7.79 / 10 | 7.51 | $6.72 | 28x more expensive |
| DeepSeek V4.1 Flash | 7.95 / 10 | 7.66 | $3.66 | 15x more expensive |
Cost breakdown
| Model | Quality | Confidence | Cost / 1k runs | Overpay | Mode |
|---|---|---|---|---|---|
| Xiaomi MiMo V2.6 Flash ★ Xiaomi MiMo | 8.17 / 10 CI [7.91, 8.44] | HIGH | $0.24 | best value | batch |
| Tencent Hy3 OpenRouter | 8.24 / 10 CI [7.96, 8.52] | HIGH | $0.77 | 3.2x | batch |
| Xiaomi MiMo V2.6 Pro Xiaomi MiMo | 8.51 / 10 CI [8.25, 8.77] | HIGH | $1.10 | 4.5x | batch |
| NVIDIA Nemotron-3 Ultra 550B OpenRouter | 8.09 / 10 CI [7.72, 8.46] | MEDIUM | $5.38 | 22x | batch |
| GLM-5.3 Flash Z.AI | 8.59 / 10 CI [8.38, 8.80] | HIGH | $15.05 | 62x | batch |
| Meta Muse Spark 1.3 OpenRouter | 8.06 / 10 CI [7.84, 8.28] | HIGH | $18.07 | 74x | batch |
| Thinking Machines Inkling OpenRouter | 8.01 / 10 CI [7.71, 8.30] | HIGH | $33.17 | 136x | batch |
| Grok 4.7 xAI | 8.28 / 10 CI [7.96, 8.61] | MEDIUM | $52.92 | 217x | batch |
| Moonshot Kimi K3 Moonshot AI | 8.48 / 10 CI [8.16, 8.81] | MEDIUM | $71.69 | 295x | batch |
| Qwen 3.8 Max Alibaba Cloud (DashScope) | 8.07 / 10 CI [7.77, 8.37] | MEDIUM | $107.70 | 443x | batch |
| GLM-5.3 best Z.AI | 8.88 / 10 CI [8.66, 9.11] | HIGH | $175.06 | 719x | 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: 3272 input tokens → 6873 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 $/1M list rates — this is what running the task costs, not a per-token price.
Evaluation rubric
Judge dependency ordering, narrative coherence, progression, local transitions, reader orientation, and use of every topic exactly once. Identifier completeness is deterministic.
Output schema
Every answer on this task is checked against this JSON Schema, whichever model wrote it. An answer that doesn't fit counts as a model failure, and the call is retried on another model.
{
"$defs": {
"TopicSequenceItem": {
"description": "Single topic in the recommended sequence.",
"properties": {
"topic_id": {
"description": "The ID of the topic (exactly as provided in input)",
"title": "Topic Id",
"type": "integer"
},
"topic_name": {
"description": "The name of the topic (exactly as provided in input, do not modify)",
"title": "Topic Name",
"type": "string"
}
},
"required": [
"topic_id",
"topic_name"
],
"title": "TopicSequenceItem",
"type": "object"
}
},
"description": "Output model for determining the optimal sequence of topics in a report.\nThe LLM analyzes topic content and determines the best reading order.",
"properties": {
"sequence_reasoning": {
"default": "",
"description": "Brief explanation of why this sequence was chosen (e.g., logical flow, building complexity, narrative arc).",
"title": "Sequence Reasoning",
"type": "string"
},
"sequenced_topics": {
"description": "Ordered list of topics in the recommended reading sequence. First topic appears first in report.",
"items": {
"$ref": "#/$defs/TopicSequenceItem"
},
"title": "Sequenced Topics",
"type": "array"
}
},
"required": [
"sequenced_topics"
],
"title": "TopicSequenceOutput",
"type": "object"
}Prompt templates
The published template for this task.
default ·
LLMB_TOPIC_SEQUENCE_OPTIMIZATION_SYSTEM +
LLMB_TOPIC_SEQUENCE_OPTIMIZATION_USER
(0 calls in window)
System prompt
Return a permutation of the supplied topics only. Place necessary foundations before dependent analysis, group related material without collapsing distinct topics, and build a deliberate opening, progression, and conclusion suited to the subject. Use topic descriptions as evidence; do not rewrite, add, drop, or duplicate topics. 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 — subject_name: {subject_name}; topic_count: {topic_count}; topics_json: {topics_json}. Use only these inputs to complete the task defined by the system prompt.