Skip to main content

Documentation Index

Fetch the complete documentation index at: https://mintlify.com/zaidmukaddam/scira/llms.txt

Use this file to discover all available pages before exploring further.

Scira supports 100+ AI models from major providers, giving you access to the latest and most powerful language models. The platform automatically selects the best model for your task, or you can manually choose based on your specific needs.

​
Model Selection

Automatic selection (default): Scira analyzes your query and selects the optimal model based on:
  • Task complexity (reasoning, generation, analysis)
  • Response time requirements
  • Cost efficiency
  • Specialized capabilities needed
Manual selection: Choose a specific model when you have requirements for:
  • Specific model capabilities (vision, function calling, extended context)
  • Consistency across sessions
  • Cost optimization
  • Provider preferences

​
Supported Providers

​
xAI

Models from X’s AI team with deep X (Twitter) integration

Grok Models

  • Grok 3: High-performance reasoning model
  • Grok 3 Mini: Faster, cost-effective variant
  • Grok 4: Latest flagship with enhanced capabilities
  • Grok 4 Fast: Optimized for speed with Grok 4 quality
  • Grok 4.1 Fast: Incremental improvements to Grok 4 Fast
  • Grok Code: Specialized for code generation and analysis
Special capabilities:
  • Native X (Twitter) search integration
  • Real-time access to X posts and trends
  • xSearch tool support
  • XQL query language (Pro)
Use cases:
  • Social media research and analysis
  • Real-time information needs
  • Code generation with Grok Code
  • General reasoning with Grok 4
Integration: Used by default in Extreme search mode and X search

​
OpenAI

Industry-leading models from OpenAI
GPT 4.1
  • GPT 4.1 Nano: Smallest variant for simple tasks
  • GPT 4.1 Mini: Balanced performance and cost
  • GPT 4.1 Standard: Full capabilities
Use cases:
  • General-purpose research
  • Content generation
  • Analysis and summarization
GPT 5
  • GPT 5 Nano: Entry-level GPT 5
  • GPT 5 Mini: Cost-effective with GPT 5 quality
  • GPT 5 Medium: Enhanced capabilities
  • GPT 5 Standard: Full GPT 5 performance
GPT 5.1
  • GPT 5.1 Instant: Fastest responses
  • GPT 5.1 Thinking: Reasoning-focused variant
  • GPT 5.1 Codex: Code generation specialist
GPT 5.2
  • GPT 5.2 Instant: Latest fast variant
  • GPT 5.2 Thinking: Enhanced reasoning
  • GPT 5.2 Codex: Latest code specialist
Use cases:
  • Advanced reasoning tasks
  • Code generation (Codex variants)
  • Quick responses (Instant variants)
  • Deep analysis (Thinking variants)
  • o3: Advanced reasoning model
  • o4 mini: Compact reasoning variant
Use cases:
  • Mathematical reasoning
  • Complex problem-solving
  • Multi-step logic
  • GPT OSS 20B: Open-source 20 billion parameter model
  • GPT OSS 120B: Open-source 120 billion parameter model
Use cases:
  • Self-hosted deployments
  • Custom fine-tuning
  • Research applications

​
Anthropic

Claude models known for safety and helpfulness
  • Claude Haiku 4.5: Fast and cost-effective
  • Claude Sonnet 4.5: Balanced performance
  • Claude Opus 4.5: Most capable Claude 4.5
Use cases:
  • Long-form content generation
  • Analysis and research
  • Creative writing
  • Safe and helpful responses
  • Claude Opus 4.6: Latest flagship model
Use cases:
  • Advanced reasoning
  • Extended context handling
  • Complex analysis

​
Google

Gemini models with multimodal capabilities
  • Gemini 2.5 Flash Lite: Fastest variant
  • Gemini 2.5 Flash Standard: Balanced Flash
  • Gemini 2.5 Pro: Full capabilities
Use cases:
  • Fast responses (Flash variants)
  • Multimodal tasks
  • General research
  • Gemini 3 Flash: Latest fast model
  • Gemini 3 Pro: Latest pro model
Use cases:
  • Cutting-edge performance
  • Advanced multimodal
  • Long context windows

​
Alibaba (Qwen)

Qwen models from Alibaba Cloud
Base Models
  • Qwen 3 4B: 4 billion parameters
  • Qwen 3 32B: 32 billion parameters
  • Qwen 3 235B: 235 billion parameters
  • Qwen 3 VL: Vision-language model
  • Qwen 3 Max: Maximum performance
  • Qwen 3 Next 80B: Next-generation 80B
Coder Models
  • Qwen 3 Coder Small: Entry-level code model
  • Qwen 3 Coder Standard: Balanced code generation
  • Qwen 3 Coder Plus: Enhanced capabilities
  • Qwen 3 Coder Next: Latest code model
Use cases:
  • Code generation (Coder variants)
  • Multilingual support
  • Vision tasks (VL variant)
  • Cost-effective deployment (smaller models)

​
Mistral

Open and commercial models from Mistral AI
  • Ministral 3 3B: 3 billion parameter compact model
  • Ministral 3 8B: 8 billion parameter model
  • Ministral 3 14B: 14 billion parameter model
Use cases:
  • Efficient local deployment
  • Cost-sensitive applications
  • Edge computing
  • Mistral Large 3: Latest large model
  • Mistral Medium: Balanced performance
  • Magistral Small: Compact variant
  • Magistral Medium: Medium variant
Use cases:
  • General research
  • European language support
  • Function calling
  • Devstral 2 Small: Entry-level code model
  • Devstral 2 Standard: Standard code generation
Use cases:
  • Code generation
  • Developer tools
  • Technical documentation

​
DeepSeek

Advanced models from DeepSeek AI
  • DeepSeek v3: Base flagship model
  • DeepSeek v3.1 Terminus: Enhanced v3.1
  • DeepSeek v3.2: Latest iteration
Use cases:
  • Deep reasoning
  • Research tasks
  • Technical analysis
  • DeepSeek R1: Reasoning-focused model
  • DeepSeek R1 0528: May 2028 release
Use cases:
  • Mathematical reasoning
  • Multi-step problems
  • Logical analysis

​
Zhipu (GLM)

GLM models from Zhipu AI
  • GLM 4.5: Base model
  • GLM 4.5 Air: Lightweight variant
  • GLM 4.6: Enhanced capabilities
  • GLM 4.6V: Vision-enabled
  • GLM 4.7: Latest flagship
  • GLM 4.7 Flash: Fast variant
Use cases:
  • Chinese language tasks
  • Multilingual support
  • Vision tasks (V variants)
  • Fast responses (Flash/Air)

​
Cohere

Enterprise-focused models from Cohere
  • Command A: Advanced reasoning
  • Command A Thinking: Reasoning-focused
Use cases:
  • Enterprise applications
  • RAG (Retrieval-Augmented Generation)
  • Embedding and reranking (used in file query tool)

​
Additional Providers

  • Kimi K2: Base model with extended context
  • Kimi K2.5: Enhanced K2
Use cases:
  • Long context handling (200K+ tokens)
  • Document analysis
  • Extended conversations
  • M1 80K: 80K context window
  • M2: Next generation
  • M2.1: Incremental update
  • M2.1 Lightning: Fast variant
Use cases:
  • Long documents
  • Fast processing (Lightning)
  • Seed 1.6: Base model
  • Seed 1.6 Flash: Fast variant
  • Seed 1.8: Latest version
Use cases:
  • General tasks
  • Cost-effective deployment
  • Trinity Mini: Compact model
  • Trinity Large: Large variant
Use cases:
  • Domain-adapted models
  • Specialized tasks
  • Vercel v0 1.0: Code generation
  • Vercel v0 1.5: Enhanced code gen
  • Amazon Nova 2 Lite: AWS model
  • Xiaomi Mimo V2 Flash: Fast model
  • StepFun Step 3.5 Flash: Reasoning model
  • Kwaipilot KAT-Coder-Pro V1: Code specialist

​
Model Capabilities

​
Function Calling

Most modern models support function calling (tool use), enabling:
  • Automatic tool selection
  • Structured output generation
  • Multi-step workflows
  • Parallel function execution
Best for function calling:
  • OpenAI GPT 4.1+, GPT 5+
  • Anthropic Claude 4.5+
  • xAI Grok 3+
  • Google Gemini 2.5+

​
Vision Capabilities

Multimodal models that process images:
  • Qwen 3 VL: Vision-language model
  • GLM 4.6V: Vision-enabled GLM
  • Google Gemini series: Native multimodal
  • OpenAI GPT-4 vision variants: Image understanding
Use cases:
  • Image analysis and description
  • OCR and text extraction
  • Visual question answering
  • Chart and diagram interpretation

​
Extended Context

Models with long context windows (100K+ tokens):
  • Kimi K2/K2.5: 200K+ tokens
  • Minimax M1 80K: 80K tokens
  • Claude 4.5/4.6: 200K tokens
  • GPT-4 variants: 128K tokens
  • Gemini Pro: 1M+ tokens
Use cases:
  • Long document analysis
  • Book summarization
  • Extended conversations
  • Large codebase understanding

​
Reasoning Models

Specialized for complex reasoning:
  • OpenAI o3, o4: Advanced reasoning
  • GPT 5.1/5.2 Thinking: Reasoning variants
  • DeepSeek R1: Reasoning-focused
  • Cohere Command A Thinking: Enterprise reasoning
  • StepFun Step 3.5: Multi-step reasoning
Use cases:
  • Mathematical proofs
  • Multi-step problems
  • Logic puzzles
  • Complex analysis

​
Code Specialists

Optimized for code generation and analysis:
  • xAI Grok Code: X-integrated code model
  • OpenAI Codex variants: GPT 5.1/5.2 Codex
  • Qwen 3 Coder series: All Coder variants
  • Mistral Devstral 2: Code generation
  • Kwaipilot KAT-Coder-Pro: Code specialist
  • Vercel v0: UI code generation
Use cases:
  • Code generation
  • Bug fixing
  • Code review
  • Documentation generation

​
Model Selection by Use Case

​
General Research

Recommended: GPT 5 Standard, Claude Sonnet 4.5, Grok 4
  • Balanced performance and cost
  • Strong reasoning capabilities
  • Good citation generation

​
Code Generation

Recommended: Grok Code, GPT 5.2 Codex, Qwen 3 Coder Plus
  • Specialized for code tasks
  • Best practices awareness
  • Multi-language support

​
Fast Responses

Recommended: GPT 5.1 Instant, Gemini 2.5 Flash, Grok 4 Fast
  • Optimized for speed
  • Lower latency
  • Good for chat interactions

​
Deep Analysis

Recommended: GPT 5.2 Thinking, Claude Opus 4.6, DeepSeek R1
  • Advanced reasoning
  • Multi-step analysis
  • Complex problem solving

​
Long Documents

Recommended: Kimi K2.5, Claude Opus 4.6, Gemini 3 Pro
  • Extended context windows
  • Document understanding
  • Cross-reference capabilities

​
Cost Optimization

Recommended: Ministral 3 8B, Qwen 3 32B, Seed 1.6 Flash
  • Open-source or cost-effective
  • Self-hosting options
  • Good performance/cost ratio

​
Social Media Research

Recommended: Grok 4, Grok 4 Fast
  • Native X integration
  • Real-time data access
  • Social context understanding

​
Provider Configuration

Scira uses the Vercel AI SDK for model integration. Supported providers are configured via: 
// From package.json dependencies:
"@ai-sdk/anthropic": "3.0.2"      // Claude models
"@ai-sdk/cohere": "3.0.2"         // Command models
"@ai-sdk/google": "3.0.2"         // Gemini models
"@ai-sdk/groq": "3.0.2"           // Groq-hosted models
"@ai-sdk/mistral": "3.0.2"        // Mistral models
"@ai-sdk/openai": "3.0.2"         // GPT models
"@ai-sdk/openai-compatible": "2.0.2" // OpenAI-compatible APIs
"@ai-sdk/xai": "3.0.3"            // Grok models
"ai": "6.0.5"                      // Vercel AI SDK core

​
Performance Considerations

Latency: Flash/Instant/Fast variants offer lower latency
  • Gemini Flash: ~200-500ms
  • GPT 5.1 Instant: ~300-600ms
  • Grok 4 Fast: ~400-700ms
Throughput: Smaller models can handle higher request volumes
  • Ministral 3B-14B: High throughput
  • Qwen 3 4B-32B: Scalable
  • Standard variants: Moderate throughput
Cost: Varies by provider and model size
  • Open models (Qwen, Mistral): Self-host cost only
  • Mini/Small variants: Lower API costs
  • Flagship models: Premium pricing

​
Next Steps

Search Modes

Learn how different modes use different models

Tools

Explore tools that leverage these models

Build docs developers (and LLMs) love

Get started for freeTalk to us