Skip to content
Guides

How to Choose the Right LLM for Each Task

JieGou supports Claude, GPT-5, and Gemini. Here's a practical guide to picking the best model for different recipe and workflow scenarios.

JT
JieGou Team
· · 3 min read

One of JieGou’s core design principles is that no single AI model is best at everything. That’s why the platform supports models from Anthropic, OpenAI, and Google — and lets you choose a different model for each recipe or workflow step.

The models available today

Anthropic Claude:

  • Claude Opus 4.5 — Most capable for complex reasoning, analysis, and nuanced writing
  • Claude Sonnet 4.5 — Strong balance of quality and speed for most tasks
  • Claude Haiku 4.5 — Fast and cost-effective for high-volume, straightforward tasks

OpenAI:

  • GPT-5.2 / GPT-5.1 — Frontier models for general-purpose tasks
  • o3 / o4-mini — Reasoning-focused models for complex analysis
  • GPT-5-mini / GPT-5-nano — Cost-optimized for simple operations

Google:

  • Gemini 3 Pro / Flash — Latest generation with strong reasoning
  • Gemini 2.5 Pro / Flash — Reliable general-purpose models
  • Gemini 2.5 Flash Lite — Budget-friendly for high-volume tasks

Matching models to tasks

Content generation and writing

Recommended: Claude Sonnet 4.5 or GPT-5.1

Writing tasks benefit from models that produce natural, nuanced prose. Claude models tend to produce well-structured long-form content. GPT-5.x models are strong at matching specific tones and styles.

Data extraction and classification

Recommended: Claude Haiku 4.5 or GPT-5-mini

Structured extraction (invoice data, ticket triage, resume screening) doesn’t need the most powerful model. Faster, cheaper models handle schema-based extraction well and keep costs low for high-volume operations.

Complex analysis and reasoning

Recommended: Claude Opus 4.5 or o3

SWOT analyses, deal risk assessment, contract clause review, and strategic planning benefit from extended thinking capabilities. Both Claude Opus and o3 support reasoning modes that let the model think through complex problems.

Summarization

Recommended: Claude Sonnet 4.5 or Gemini 2.5 Flash

Summarization is a well-understood task where most models perform comparably. Choose based on speed and cost rather than raw capability.

Per-step model selection in workflows

In a workflow, you can assign different models to each step. For example:

  1. Data extraction (Haiku 4.5) — Fast, cheap extraction of invoice data
  2. Discrepancy analysis (Sonnet 4.5) — Moderate reasoning to flag mismatches
  3. Approval summary (Haiku 4.5) — Simple formatting for the approval notification
  4. Final report (Opus 4.5) — High-quality narrative for stakeholders

This approach optimizes both cost and quality — you use premium models only where they make a measurable difference.

Bring your own keys

JieGou doesn’t proxy your API calls or add markup. You connect your own API keys for each provider, and token usage flows directly to your provider accounts. The platform tracks usage so you can see costs per recipe, workflow, and department.

This means you have full control over rate limits, spending caps, and data handling policies — everything is between you and the provider.

Web search and extended thinking

Some models support additional capabilities:

  • Web search is available on all Claude models, GPT-5.x, and Gemini 2.5+
  • Extended thinking is available on all Claude models, o3/o4-mini, and Gemini 3

These features are toggled at the recipe level, so you can enable web search for prospect research but disable it for internal document processing.

Start experimenting

The simplest way to compare models is to create a recipe and run it with different models using the same inputs. JieGou tracks execution time, token counts, and output quality feedback, making it easy to find the best fit for each task.

llm models anthropic openai google
Share this article

Enjoyed this post?

Get workflow tips, product updates, and automation guides in your inbox.

No spam. Unsubscribe anytime.