Getting Started

Model Discovery Guide

LangMart makes it easy to discover the perfect AI model for your needs. This guide covers advanced browsing, comparing models, and understanding model capabilities.

Featured models are curated selections highlighted for:

  • Performance - Best-in-class results
  • Value - Great quality for the price
  • Novelty - New and noteworthy releases
  • Popularity - Most-used by the community

Featured models appear in several places:

  1. Homepage - Quick access to top models
  2. Models Page - Featured tab (when available)
  3. Chat Selector - Promoted in model picker

Models may be featured in categories:

Category Criteria
Top Picks Editor's choice for quality
Best Value Quality vs. price ratio
Fastest Lowest latency models
Newest Recent releases
Most Popular Highest usage

Browsing by Provider

Provider Tabs

The Models page organizes models by provider:

Navigation:

  1. Go to the Models page
  2. Click a provider tab to filter
  3. All models from that provider appear
  4. Use additional filters within the provider view

Provider-Specific Features

Each provider offers unique capabilities:

OpenAI:

  • GPT-4 series with vision
  • Function calling support
  • JSON mode output
  • Structured outputs

Anthropic:

  • Claude 3.5 with 200K context
  • Strong reasoning abilities
  • Built-in safety features
  • Excellent code generation

Google:

  • Gemini with multimodal input
  • Up to 1M token context
  • Fast inference options
  • Native Google integration

Meta (Llama):

  • Open weights models
  • Available via multiple hosts
  • Customizable deployments
  • Cost-effective options

Mistral:

  • Efficient architecture
  • Open and commercial options
  • Strong multilingual support
  • Mixture of Experts models

Provider Badges

Model cards show provider-specific badges:

  • Provider logo/icon
  • Model family indicator
  • Version information
  • Special features

Comparing Models

Side-by-Side Comparison

Compare models to find the best fit:

  1. Open model details for a model
  2. Click Compare or add to comparison
  3. Select additional models
  4. View comparison table

Comparison Criteria

Compare models across key dimensions:

Dimension What to Compare
Pricing Input/output costs per token
Context Maximum context window
Speed Typical response latency
Capabilities Features supported
Quality Output quality ratings

Making Trade-offs

Consider these trade-offs:

Quality vs. Cost:

  • Premium models cost more but deliver better results
  • Mini/Flash variants offer good quality at lower cost
  • Use premium for important tasks, budget models for simple ones

Speed vs. Quality:

  • Faster models may sacrifice some quality
  • Groq and Flash models prioritize speed
  • Full models take longer but may be more thorough

Context vs. Cost:

  • Larger context windows cost more per request
  • Only use large context when needed
  • Consider summarization for very long inputs

Benchmark Comparisons

Some models include benchmark scores:

  • MMLU - Multi-task language understanding
  • HumanEval - Code generation accuracy
  • Math - Mathematical reasoning
  • Reasoning - Logical problem solving

Model Capabilities

Understanding Capability Badges

Model cards display capability badges:

Badge Meaning
Vision Can process images
Reasoning Enhanced thinking/analysis
Tool Use Can call functions
Image Gen Creates images
Audio Processes audio input
Embedding Creates vector embeddings
JSON Mode Structured JSON output
Streaming Real-time token streaming

Vision Capability

Models with vision can:

  • Analyze images you upload
  • Read text from images (OCR)
  • Describe visual content
  • Answer questions about images

Using Vision:

  1. Select a vision-capable model
  2. Attach an image to your message
  3. Ask about the image content

Reasoning Capability

Reasoning models excel at:

  • Complex problem solving
  • Multi-step analysis
  • Logical deduction
  • Mathematical reasoning

Best for:

  • Technical problems
  • Data analysis
  • Strategic planning
  • Code debugging

Tool Use Capability

Tool-capable models can:

  • Call functions you define
  • Use built-in tools (in Remote Chat)
  • Chain multiple tool calls
  • Handle complex workflows

Requirement: Enable tools in Remote Chat mode

Context Windows

Context window determines input capacity:

Small (4K-8K tokens):

  • Short conversations
  • Simple queries
  • Quick responses

Medium (32K-64K tokens):

  • Document analysis
  • Extended conversations
  • Code review

Large (128K-200K tokens):

  • Long documents
  • Multi-file analysis
  • Book-length content

Very Large (1M+ tokens):

  • Entire codebases
  • Research papers
  • Comprehensive analysis

Streaming Support

Most models support streaming:

  • Text appears word by word
  • Faster perceived response time
  • Can stop generation early
  • Better user experience

Advanced Discovery

Search Strategies

By Name:

Search: "gpt-4" "claude" "llama"

By Capability:

Filter: Vision + Reasoning

By Provider + Capability:

Tab: OpenAI
Filter: Tool Use

Finding Specific Models

Latest Models:

  • Sort by release date
  • Check "Newest" featured section
  • Follow provider announcements

Budget Models:

  • Filter by billing (Self-paid)
  • Sort by price low-to-high
  • Look for "mini" or "flash" variants

Enterprise Models:

  • Filter by billing (Org-paid)
  • Check security features
  • Verify compliance certifications

Model Versioning

Models come in versions:

Naming Patterns:

  • gpt-4-0613 - Date-versioned
  • claude-3-sonnet - Named versions
  • llama-3.1-70b - Size and version

Version Considerations:

  • Newer isn't always better for your use case
  • Date versions are more stable
  • Latest may have improvements or changes

Model Selection Workflow

Step 1: Define Requirements

Ask yourself:

  1. What task am I doing?
  2. What capabilities do I need?
  3. What's my budget?
  4. How important is speed?
  5. How much context do I need?

Step 2: Filter Candidates

Use filters to narrow down:

  1. Set capability requirements
  2. Filter by billing mode
  3. Consider context needs
  4. Note pricing range

Step 3: Compare Options

For top candidates:

  1. Review detailed specifications
  2. Compare pricing
  3. Check benchmark scores
  4. Read descriptions

Step 4: Test and Iterate

Try models in practice:

  1. Use each for sample tasks
  2. Evaluate response quality
  3. Note response speed
  4. Check cost tracking

Step 5: Organize Favorites

Once you find winners:

  1. Add to favorites
  2. Create collections for different tasks
  3. Set up quick access

Tips for Discovery

Stay Updated

  • Check featured models regularly
  • New models are added frequently
  • Pricing and capabilities change
  • Follow LangMart announcements

Use Multiple Models

  • Different models for different tasks
  • Collections for A/B testing
  • Fallback options for availability

Consider Total Cost

  • Input and output pricing differ
  • Context window affects cost
  • Batch small requests
  • Monitor usage in analytics

Next Steps