The Crescitaly benchmark for growth teams. Compare frontier AI models on realistic campaign rescue, UGC scripting, SEO execution, localization, and delivery-ops tasks to find the best operator for your workflow.
Static snapshot fallback while the live benchmark is not yet configured.
Snapshot
Pilot public benchmark • March 21, 2026, 11:20 AM GST
Models
10
Top Score
95.8
Last Indexed
March 21, 2026, 11:20 AM GST
Public benchmark snapshot
The Crescitaly benchmark for growth teams. Compare frontier AI models on strategy, hooks, creative, SEO, localization, and delivery ops. Scores are normalized to 0-100 across Crescitaly-style evaluations. This v0 snapshot is designed to be replaced by live evaluator runs.
| Rank | Model | Overall | Strat | Hooks | Creative | SEO | Comp % | Cost | Speed |
|---|---|---|---|---|---|---|---|---|---|
| 1 | OPGPT-5.4 openai/gpt-5.4 | 95.8 | 98.2 | 95.1 | 93.6 | 97.4 | 96.8% | $4.9 | 41.8s |
| 2 | CLClaude Sonnet 4.6 anthropic/claude-sonnet-4.6 | 95.1 | 96.8 | 94.6 | 94.1 | 96.2 | 95.9% | $5.4 | 37.2s |
| 3 | CLClaude Opus 4.6 anthropic/claude-opus-4.6 | 94.9 | 96.3 | 95.5 | 95.8 | 95.4 | 95.4% | $7.1 | 44.6s |
| 4 | XAGrok 4.20 Multi-Agent x-ai/grok-4.20-multi-agent | 94.6 | 95.4 | 96.8 | 96.1 | 92.1 | 94.7% | $6.3 | 58.3s |
| 5 | OPGPT-5.4 Mini openai/gpt-5.4-mini | 94.4 | 95.2 | 92.9 | 91.1 | 95.7 | 97.3% | $1.2 | 18.4s |
| 6 | OPGPT-5.3 Codex openai/gpt-5.3-codex | 93.8 | 95.0 | 91.2 | 89.7 | 95.8 | 93.9% | $3.7 | 46.2s |
| 7 | GGGemini 3.1 Pro Preview google/gemini-3.1-pro-preview | 92.9 | 94.1 | 90.5 | 91.2 | 94.6 | 92.8% | $5.8 | 49.1s |
| 8 | QWQwen 3.5 Plus 02-15 qwen/qwen-3.5-plus-02-15 | 90.8 | 92.6 | 88.4 | 86.9 | 91.7 | 90.4% | $2.6 | 63.4s |
| 9 | ZAGLM-5 z-ai/glm-5 | 88.9 | 90.2 | 84.7 | 82.8 | 89.1 | 88.6% | $2.1 | 57.9s |
| 10 | KMKimi K2.5 moonshot/kimi-k2.5 | 87.6 | 89.4 | 87.1 | 88.6 | 84.3 | 86.2% | $1.9 | 71.6s |
Benchmark 01
Audit an underperforming service landing page, identify the three biggest conversion leaks, rewrite the hero stack, propose two proof sections, and hand over a short ad angle map for paid social.
Deliverable
Rescue memo with hero, CTA, FAQ rewrites, and a paid-social angle table.
Judge Focus
We reward diagnosis depth, conversion logic, and how ready the handoff is for a marketer and designer.
Sorted By
Overall Score · highest first
OpenAI GPT-5.4
Top compare card #1
97.2
44.1s
Produced the most balanced rewrite between CRO, clarity, and operator-ready next steps.
Best At
Strongest end-to-end rescue memo and the cleanest stakeholder handoff.
Watch For
Can over-polish sections that only needed a lighter edit.
Anthropic Claude Sonnet 4.6
Top compare card #2
96.8
36.8s
Best when the brief needs polish, trust-building, and multilingual nuance.
Best At
Excellent objection handling and sharper FAQ restructuring.
Watch For
Softer on aggressive CTA experimentation.
xAI Grok 4.20 Multi-Agent
Top compare card #3
95.9
58.7s
Useful when you want more swing and more testing directions in one pass.
Best At
Generated the widest angle map and strongest diagnostic depth.
Watch For
Needs trimming to avoid over-delivering beyond the ask.
OpenAI GPT-5.4 Mini
Top compare card #4
94.6
19.2s
The best speed-to-value option for triage and first-pass iteration.
Best At
Fastest usable first draft with very strong structure.
Watch For
Less persuasive detail in proof blocks than the larger frontier models.
The scenarios are based on actual Crescitaly work: campaign rescue briefs, UGC scripting, multilingual landing rewrites, SEO clusters, and operator handoffs.
Each run is graded on clarity, conversion intent, structural completeness, localization fidelity, and how close the output is to something a team can ship.
This public page is seeded with a pilot snapshot and a reusable data model, so it can be swapped to real evaluator outputs without redesigning the surface.
Next Step
The public surface is ready. The next step is wiring your preferred model APIs, storing run artifacts, and replacing the pilot snapshot with repeatable benchmark executions.