/**
* Groq Model Specifications
* Updated June 17, 2025 with latest Groq model releases
*
* Groq provides ultra-fast inference with OpenAI-compatible API
* Specializes in high-performance LLM serving with extended usage metrics
*/
import { ModelSpec } from '../modelTypes';
export const GROQ_MODELS: ModelSpec[] = [
// Current Production Models (January 2025)
// Llama 3.1 models - Text generation
{
provider: 'groq',
name: 'Llama 3.1 70B Versatile',
apiName: 'llama-3.1-70b-versatile',
contextWindow: 131072,
maxTokens: 8192,
inputCostPerMillion: 0.59,
outputCostPerMillion: 0.79,
capabilities: {
supportsJSON: true,
supportsImages: false,
supportsFunctions: true,
supportsStreaming: true,
supportsThinking: false
}
},
{
provider: 'groq',
name: 'Llama 3.1 8B Instant',
apiName: 'llama-3.1-8b-instant',
contextWindow: 131072,
maxTokens: 8192,
inputCostPerMillion: 0.05,
outputCostPerMillion: 0.08,
capabilities: {
supportsJSON: true,
supportsImages: false,
supportsFunctions: true,
supportsStreaming: true,
supportsThinking: false
}
},
// Llama 3.3 models
{
provider: 'groq',
name: 'Llama 3.3 70B Versatile',
apiName: 'llama-3.3-70b-versatile',
contextWindow: 128000,
maxTokens: 32768,
inputCostPerMillion: 0.59,
outputCostPerMillion: 0.79,
capabilities: {
supportsJSON: true,
supportsImages: false,
supportsFunctions: true,
supportsStreaming: true,
supportsThinking: false
}
},
// Gemma models
{
provider: 'groq',
name: 'Gemma 2 9B IT',
apiName: 'gemma2-9b-it',
contextWindow: 8192,
maxTokens: 8192,
inputCostPerMillion: 0.20,
outputCostPerMillion: 0.20,
capabilities: {
supportsJSON: true,
supportsImages: false,
supportsFunctions: true,
supportsStreaming: true,
supportsThinking: false
}
},
// New Llama 4 models
{
provider: 'groq',
name: 'Llama 4 Scout',
apiName: 'llama-4-scout-17bx16e',
contextWindow: 128000,
maxTokens: 8192,
inputCostPerMillion: 0.11,
outputCostPerMillion: 0.34,
capabilities: {
supportsJSON: true,
supportsImages: false,
supportsFunctions: true,
supportsStreaming: true,
supportsThinking: false
}
},
{
provider: 'groq',
name: 'Llama 4 Maverick',
apiName: 'llama-4-maverick-17bx128e',
contextWindow: 128000,
maxTokens: 8192,
inputCostPerMillion: 0.20,
outputCostPerMillion: 0.60,
capabilities: {
supportsJSON: true,
supportsImages: false,
supportsFunctions: true,
supportsStreaming: true,
supportsThinking: false
}
},
// Content moderation model
{
provider: 'groq',
name: 'Llama Guard 4 12B',
apiName: 'meta-llama/llama-guard-4-12b',
contextWindow: 131072,
maxTokens: 128,
inputCostPerMillion: 0.20,
outputCostPerMillion: 0.20,
capabilities: {
supportsJSON: true,
supportsImages: false,
supportsFunctions: false,
supportsStreaming: true,
supportsThinking: false
}
},
// OpenAI models on Groq
{
provider: 'groq',
name: 'GPT OSS 20B',
apiName: 'openai/gpt-oss-20b',
contextWindow: 131072,
maxTokens: 32768,
inputCostPerMillion: 0.10,
outputCostPerMillion: 0.50,
capabilities: {
supportsJSON: true,
supportsImages: false,
supportsFunctions: true,
supportsStreaming: true,
supportsThinking: true
}
},
{
provider: 'groq',
name: 'GPT OSS 120B',
apiName: 'openai/gpt-oss-120b',
contextWindow: 131072,
maxTokens: 32766,
inputCostPerMillion: 0.149,
outputCostPerMillion: 0.769,
capabilities: {
supportsJSON: true,
supportsImages: false,
supportsFunctions: true,
supportsStreaming: true,
supportsThinking: true
}
},
// Preview models
{
provider: 'groq',
name: 'DeepSeek R1 Distill Llama 70B',
apiName: 'deepseek-r1-distill-llama-70b',
contextWindow: 128000,
maxTokens: 8192,
inputCostPerMillion: 0.75,
outputCostPerMillion: 0.99,
capabilities: {
supportsJSON: true,
supportsImages: false,
supportsFunctions: true,
supportsStreaming: true,
supportsThinking: true
}
},
{
provider: 'groq',
name: 'Qwen 3 32B',
apiName: 'qwen/qwen3-32b',
contextWindow: 128000,
maxTokens: 16384,
inputCostPerMillion: 0.29,
outputCostPerMillion: 0.59,
capabilities: {
supportsJSON: true,
supportsImages: false,
supportsFunctions: true,
supportsStreaming: true,
supportsThinking: false
}
}
];
export const GROQ_DEFAULT_MODEL = 'llama-3.1-70b-versatile';
/**
* Groq-specific model categories for easier selection
*/
export const GROQ_MODEL_CATEGORIES = {
// Ultra-fast text generation
FAST_TEXT: [
'llama-3.1-8b-instant',
'gemma2-9b-it'
],
// High-quality text generation
QUALITY_TEXT: [
'llama-3.1-70b-versatile',
'llama-3.3-70b-versatile',
'llama-4-maverick-17bx128e'
],
// New Llama 4 models
LLAMA_4: [
'llama-4-scout-17bx16e',
'llama-4-maverick-17bx128e'
],
// Reasoning-optimized
REASONING: [
'deepseek-r1-distill-llama-70b'
],
// Content moderation
MODERATION: [
'meta-llama/llama-guard-4-12b'
],
// Preview/experimental models
PREVIEW: [
'deepseek-r1-distill-llama-70b',
'qwen/qwen3-32b'
]
};
/**
* Get models by category
*/
export function getGroqModelsByCategory(category: keyof typeof GROQ_MODEL_CATEGORIES): ModelSpec[] {
const modelNames = GROQ_MODEL_CATEGORIES[category];
return GROQ_MODELS.filter(model => modelNames.includes(model.apiName));
}
/**
* Check if a model supports specific capabilities
*/
export function getGroqModelCapabilities(modelName: string): ModelSpec['capabilities'] | null {
const model = GROQ_MODELS.find(m => m.apiName === modelName);
return model?.capabilities || null;
}
