107 lines · 3.1 KB · TypeScript
RawDownload
import type { StudioModelRecommendation, StudioModuleDefaults } from "./types"; export const STUDIO_MODULE_DEFAULTS: StudioModuleDefaults = {  uiRefreshMs: 5_000,}; export const STUDIO_MODEL_RECOMMENDATIONS: StudioModelRecommendation[] = [  // --- Flagship / large ---  {    id: "meta-llama/Llama-4-Maverick-17B-128E-Instruct",    name: "Llama 4 Maverick 17Bx128E",    size_gb: 160,    min_vram_gb: 140,    description: "Meta's latest MoE flagship — 400B+ params, top-tier reasoning.",    tags: ["chat", "reasoning", "flagship"],  },  {    id: "Qwen/Qwen3-235B-A22B",    name: "Qwen3 235B (A22B MoE)",    size_gb: 150,    min_vram_gb: 130,    description: "Qwen's largest MoE model with thinking/non-thinking modes.",    tags: ["chat", "reasoning", "multilingual", "flagship"],  },  {    id: "deepseek-ai/DeepSeek-R1",    name: "DeepSeek R1",    size_gb: 160,    min_vram_gb: 140,    description: "Top-tier open reasoning model, 671B MoE.",    tags: ["reasoning", "code", "flagship"],  },  // --- High-quality mid-range ---  {    id: "meta-llama/Llama-4-Scout-17B-16E-Instruct",    name: "Llama 4 Scout 17Bx16E",    size_gb: 38,    min_vram_gb: 34,    description: "Efficient Llama 4 MoE variant with 10M token context.",    tags: ["chat", "long-context", "recommended"],  },  {    id: "Qwen/Qwen3-32B",    name: "Qwen3 32B",    size_gb: 64,    min_vram_gb: 48,    description: "Dense 32B with built-in thinking — strong all-rounder.",    tags: ["chat", "reasoning", "code", "recommended"],  },  {    id: "meta-llama/Llama-3.3-70B-Instruct",    name: "Llama 3.3 70B Instruct",    size_gb: 140,    min_vram_gb: 80,    description: "Latest Llama 3.3 dense model, excellent instruction following.",    tags: ["chat", "general", "recommended"],  },  {    id: "mistralai/Mistral-Small-24B-Instruct-2501",    name: "Mistral Small 24B",    size_gb: 48,    min_vram_gb: 32,    description: "Mistral's latest efficient model with strong tool use.",    tags: ["chat", "tool-use", "fast"],  },  {    id: "google/gemma-3-27b-it",    name: "Gemma 3 27B",    size_gb: 54,    min_vram_gb: 40,    description: "Google's latest open model — multilingual, vision-ready.",    tags: ["chat", "multilingual", "vision"],  },  // --- Compact / efficient ---  {    id: "Qwen/Qwen3-14B",    name: "Qwen3 14B",    size_gb: 28,    min_vram_gb: 20,    description: "Great quality-to-size ratio with thinking mode.",    tags: ["chat", "reasoning", "fast"],  },  {    id: "meta-llama/Llama-3.1-8B-Instruct",    name: "Llama 3.1 8B",    size_gb: 16,    min_vram_gb: 12,    description: "Fast and reliable for single-GPU setups.",    tags: ["chat", "fast", "starter"],  },  {    id: "Qwen/Qwen3-8B",    name: "Qwen3 8B",    size_gb: 16,    min_vram_gb: 12,    description: "Compact model with thinking and tool-use support.",    tags: ["chat", "reasoning", "fast", "starter"],  },  {    id: "microsoft/Phi-4",    name: "Phi-4 14B",    size_gb: 28,    min_vram_gb: 20,    description: "Microsoft's latest small-but-capable reasoning model.",    tags: ["chat", "reasoning", "code"],  },];