Best LLMs for Promotional Campaign Generation

Category: Social & Promotional Content · Typical I/O: 3835→2691 tokens

Runs on model group: All models. Inputs may reach any model in the group, not only the ranked models below.

Run this task on Fronset — request an invitation

Models

Frontier on this task: Claude Opus 5.5 at 7.98 / 10. Quality bar at 90%: 7.19.

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
Claude Haiku 5.5	7.45 / 10	7.07	$1.77	best value
MiniMax M3	7.94 / 10	7.44	$2.56	1.4x more expensive
GPT-6 Sol	7.84 / 10	7.49	$18.57	10x more expensive
Claude Sonnet 5.5	7.78 / 10	7.47	$30.70	17x more expensive
Claude Opus 5.5	7.98 / 10	7.69	$65.07	37x more expensive
Grok 4.7	7.51 / 10	7.21	$227.70	129x more expensive
Moonshot Kimi K3	7.78 / 10	7.50	$595.24	337x more expensive
Qwen 3.8 Flash	5.97 / 10	5.47	$48.67	28x more expensive
Tencent Hy3	6.81 / 10	6.37	$2.55	1.4x more expensive
Qwen 3.8 Max	6.85 / 10	6.48	$324.61	184x more expensive
Gemini 3.8 Flash	6.49 / 10	6.10	$44.12	25x more expensive
Xiaomi MiMo V2.6 Flash	6.48 / 10	6.06	$2.10	1.2x more expensive
DeepSeek V4.1 Flash	7.15 / 10	6.85	$11.49	6.5x more expensive

Cost breakdown

Model	Quality	Confidence	Cost / 1k runs	Overpay	Mode
Claude Haiku 5.5 ★ Anthropic	7.45 / 10 CI [7.07, 7.82]	MEDIUM	$1.77	best value	batch
MiniMax M3 OpenRouter	7.94 / 10 CI [7.44, 8.43]	MEDIUM	$2.56	1.4x	batch
GPT-6 Sol OpenAI	7.84 / 10 CI [7.49, 8.20]	MEDIUM	$18.57	10x	batch
Claude Sonnet 5.5 Anthropic	7.78 / 10 CI [7.47, 8.09]	MEDIUM	$30.70	17x	batch
Claude Opus 5.5 best Anthropic	7.98 / 10 CI [7.69, 8.28]	HIGH	$65.07	37x	batch
Grok 4.7 xAI	7.51 / 10 CI [7.21, 7.80]	HIGH	$227.70	129x	batch
Moonshot Kimi K3 Moonshot AI	7.78 / 10 CI [7.50, 8.05]	HIGH	$595.24	337x	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: 3835 input tokens → 2691 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 campaign-level variety and progression, per-message grounding, phase and timing fit, compliance with the supplied channel profile, hook quality without hype, non-duplication, and faithful handling of release status. Deterministic limits are out of scope.

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.

{
  "$defs": {
    "XcomPostItem": {
      "description": "Single X.com post proposal.",
      "properties": {
        "content": {
          "description": "The post text (under 280 chars for non-threads)",
          "title": "Content",
          "type": "string"
        },
        "post_type": {
          "description": "Type of post: THREAD_STARTER, SINGLE_INSIGHT, DATA_POINT, QUESTION, or HOT_TAKE",
          "title": "Post Type",
          "type": "string"
        }
      },
      "required": [
        "post_type",
        "content"
      ],
      "title": "XcomPostItem",
      "type": "object"
    }
  },
  "description": "Output model for X.com post proposals generation.\nGenerates multiple post variants optimized for social media engagement.",
  "properties": {
    "posts": {
      "description": "List of 5 X.com post proposals",
      "items": {
        "$ref": "#/$defs/XcomPostItem"
      },
      "title": "Posts",
      "type": "array"
    }
  },
  "required": [
    "posts"
  ],
  "title": "XcomProposalsOutput",
  "type": "object"
}

Prompt templates

The published template for this task.

default · LLMB_PROMOTIONAL_CAMPAIGN_GENERATION_SYSTEM + LLMB_PROMOTIONAL_CAMPAIGN_GENERATION_USER (0 calls in window)

System prompt

Build a varied campaign whose messages each express one source-grounded idea and whose sequence develops rather than repeats the same hook. Apply channel_profile and campaign_profile as the sole source of channel, message format, phases, release behavior, links, cadence, and prohibited tactics. Avoid invented urgency, unsupported performance claims, repetitive templates, and unverified news framing. 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 — post_count: {post_count}; title_context: {title_context}; exec_summary_section: {exec_summary_section}; content: {content}; post_types: {post_types}; datetime_now: {datetime_now}; published_on: {published_on}; synthesis_report: {synthesis_report}; channel_profile: {channel_profile}; campaign_profile: {campaign_profile}. Use only these inputs to complete the task defined by the system prompt.