Skip to content

Qwen3.8-Flash-Next on any consumer hardware: one-click install for Windows / Linux. Strata inference engine, OpenAI/Anthropic API on localhost, optional image input.

C++
8.7 KB
Raw Blame
{
  "available": true,
  "version": "v0.1.38",
  "repo": "/Applications/Strata/engine",
  "summary": "UI preview — simulated model state",
  "error": null,
  "catalog": {
    "families": [
      {
        "family": "qwen",
        "title": "Qwen3.8-Flash-Next",
        "about": "the original model",
        "experimental": false,
        "images": true,
        "sizes": [
          {
            "id": "q2_0",
            "model": "Q2_0",
            "about": "2-bit, the fastest",
            "download_gb": 66.4,
            "ram_needed_gb": 48,
            "experts_gb": 34,
            "on_this_pc": "does not fit",
            "installed": false
          },
          {
            "id": "iq2_xs",
            "model": "IQ2_XS",
            "about": "2-bit i-quant, a little better quality, close in speed",
            "download_gb": 68,
            "ram_needed_gb": 48,
            "experts_gb": 35.5,
            "on_this_pc": "does not fit",
            "installed": false
          },
          {
            "id": "iq3_xxs",
            "model": "IQ3_XXS",
            "about": "3-bit i-quant, better quality, slower (more CPU work per token)",
            "download_gb": 75.8,
            "ram_needed_gb": 60,
            "experts_gb": 42.9,
            "on_this_pc": "does not fit",
            "installed": false
          },
          {
            "id": "iq3_s",
            "model": "IQ3_S",
            "about": "3.5-bit i-quant, the best quality (matches the full model), the slowest; needs a 64 GB PC with little else running",
            "download_gb": 83.6,
            "ram_needed_gb": 62,
            "experts_gb": 50.3,
            "on_this_pc": "does not fit",
            "installed": false
          }
        ]
      },
      {
        "family": "swift",
        "title": "Swift 1.5",
        "about": "thinks much shorter (-63% thinking tokens, 1.8x sooner answers by its authors' numbers)",
        "experimental": false,
        "images": true,
        "sizes": [
          {
            "id": "swift-iq2_xs",
            "model": "IQ2_XS",
            "about": "2-bit i-quant, a little better quality, close in speed",
            "download_gb": 68,
            "ram_needed_gb": 48,
            "experts_gb": 35.5,
            "on_this_pc": "does not fit",
            "installed": false
          },
          {
            "id": "swift-iq3_xxs",
            "model": "IQ3_XXS",
            "about": "3-bit i-quant, better quality, slower (more CPU work per token)",
            "download_gb": 75.8,
            "ram_needed_gb": 60,
            "experts_gb": 42.9,
            "on_this_pc": "does not fit",
            "installed": false
          }
        ]
      },
      {
        "family": "coder",
        "title": "Qwen3.8-Flash-Next Coder",
        "about": "half the experts (code, tools, images kept): needs ~32 GB of RAM, faster; weaker outside coding",
        "experimental": false,
        "images": true,
        "sizes": [
          {
            "id": "coder-iq1_m",
            "model": "IQ1_M",
            "about": "the Coder's only size: half the experts, stored like IQ3_S (3.5 bits)",
            "download_gb": 58.4,
            "ram_needed_gb": 32,
            "experts_gb": 23.4,
            "on_this_pc": "does not fit",
            "installed": false
          }
        ]
      },
      {
        "family": "unsloth",
        "title": "Qwen3.8-Flash-Next (Unsloth)",
        "about": "4-bit, 111 GB download, most experts read from the SSD: slow (7-8.5 tokens/s on a 64 GB PC)",
        "experimental": true,
        "images": false,
        "sizes": [
          {
            "id": "unsloth-ud-q4_k_xl",
            "model": "UD-Q4_K_XL",
            "about": "4-bit (Unsloth Dynamic), EXPERIMENTAL: the best quality, but most experts come from the SSD on a 64 GB PC (7-8.5 tokens/s measured)",
            "download_gb": 111.3,
            "ram_needed_gb": 48,
            "experts_gb": 77,
            "on_this_pc": "does not fit",
            "installed": false
          }
        ]
      }
    ],
    "contexts": [
      8192,
      32768,
      65536,
      131072,
      262144,
      393216,
      524288
    ],
    "recommendation": {
      "family": "qwen",
      "model": "Q2_0",
      "context": 32768
    },
    "context_note": "longer contexts need more VRAM (fewer experts fit on the GPU); past 262144 setup adds rope scaling (experimental)"
  },
  "fields": [
    {
      "id": "reasoning_budget_tokens",
      "label": "Thinking token budget",
      "group": "Responses",
      "type": "number",
      "min": 0,
      "max": 1048576,
      "hint": "Leave blank for the model default.",
      "supported": true
    },
    {
      "id": "fit_max_tokens",
      "label": "Fit output to remaining context",
      "group": "Responses",
      "type": "boolean",
      "hint": "Shortens the requested output cap when the prompt leaves less room.",
      "supported": true
    },
    {
      "id": "anthropic_thinking",
      "label": "Anthropic thinking",
      "group": "Responses",
      "type": "enum",
      "choices": [
        "model",
        "on_request"
      ],
      "hint": "Follow the model default or think only when requested.",
      "supported": true
    },
    {
      "id": "effort_position",
      "label": "Reasoning effort position",
      "group": "Responses",
      "type": "enum",
      "choices": [
        "start",
        "end"
      ],
      "supported": false
    },
    {
      "id": "aliases",
      "label": "API model aliases",
      "group": "Serving",
      "type": "array",
      "hint": "Comma-separated names clients may use for this model.",
      "supported": true
    },
    {
      "id": "idle_unload_s",
      "label": "Unload after idle seconds",
      "group": "Serving",
      "type": "number",
      "min": 0,
      "max": 86400,
      "hint": "Zero keeps the model loaded.",
      "supported": true
    },
    {
      "id": "lazy_load",
      "label": "Load on first request",
      "group": "Serving",
      "type": "boolean",
      "supported": true
    },
    {
      "id": "parallel",
      "label": "Concurrent requests",
      "group": "Serving",
      "type": "number",
      "min": 1,
      "max": 8,
      "hint": "More slots need more memory and can slow each response.",
      "supported": false
    },
    {
      "id": "engine_silence_s",
      "label": "Engine stall timeout (seconds)",
      "group": "Recovery",
      "type": "number",
      "min": 10,
      "max": 3600,
      "supported": true
    },
    {
      "id": "api_monitor",
      "label": "Keep API request metadata",
      "group": "Recovery",
      "type": "boolean",
      "supported": true
    },
    {
      "id": "layer_split",
      "label": "Layer split",
      "group": "GPU & memory",
      "type": "string",
      "hint": "auto, or increasing split points such as 18 or 16,32.",
      "supported": true
    },
    {
      "id": "gpu_order",
      "label": "GPU order",
      "group": "GPU & memory",
      "type": "enum",
      "choices": [
        "auto",
        "as_given"
      ],
      "supported": false
    },
    {
      "id": "split_skip_if_fits",
      "label": "Use one GPU when the model fits",
      "group": "GPU & memory",
      "type": "boolean",
      "supported": true
    },
    {
      "id": "vram_elastic",
      "label": "Elastic GPU expert cache",
      "group": "GPU & memory",
      "type": "boolean",
      "supported": false
    },
    {
      "id": "vram_reserve_mib",
      "label": "GPU memory for other apps (MiB)",
      "group": "GPU & memory",
      "type": "number",
      "min": 0,
      "max": 131072,
      "hint": "Applied as --vram-reserve-mib when loading.",
      "supported": true
    },
    {
      "id": "vision_tokens",
      "label": "Maximum image tokens",
      "group": "Image reading",
      "type": "number",
      "min": 1,
      "max": 16384,
      "supported": true
    },
    {
      "id": "vision_min_tokens",
      "label": "Minimum image tokens",
      "group": "Image reading",
      "type": "number",
      "min": 0,
      "max": 16384,
      "supported": true
    },
    {
      "id": "mcp_timeout_s",
      "label": "Tool timeout (seconds)",
      "group": "Tools",
      "type": "number",
      "min": 1,
      "max": 3600,
      "supported": true
    },
    {
      "id": "mcp_max_rounds",
      "label": "Maximum tool rounds",
      "group": "Tools",
      "type": "number",
      "min": 1,
      "max": 100,
      "supported": true
    },
    {
      "id": "mcp_max_result_chars",
      "label": "Maximum tool result characters",
      "group": "Tools",
      "type": "number",
      "min": 100,
      "max": 1000000,
      "supported": true
    }
  ],
  "capabilities": {
    "updates": false,
    "rollback": false,
    "calibration": true,
    "responses": true
  },
  "instances": [],
  "jobs": [],
  "defaultId": null
}