Inference · season 2026-W46

Season 2026-W46, open.

9 Nov 2026 to 16 Nov 2026 — 7 days left. What the grams proved of the catalog this window, per artifact and hardware class. Lineage is from the catalog; every number is a signed measurement or a count of them, and a gram count is a verified floor.

0grams contributed
0artifacts proved
0hardware classes
0publishers
cohorts this season
  • 0 · Lead member, rung 00
  • 1 · Lead member's ladder0
  • 2 · Other vendor members0
  • 3 · Derivatives0

The order a season proves things in: the number every other number is compared to comes first.

hardware classes that signed

    A class is os · arch · card class · RAM band. A number proved on one says nothing about another.

    cells the network has signed as unprovable
    • myshell-ai/OpenVoiceV2Linux · x86-64 · no GPU · 32 GiB · unloadable · 3 grams

      the engine has no loader for this architecture — a release changes that, never a clock

    • hf.co/huihui-ai/Huihui-Qwen3.6-35B-A3B-abliterated-MTP-GGUF:Q3_KLinux · x86-64 · no GPU · 32 GiB · unavailable · 2 grams

      the ref, or the quant asked of it, is not published upstream

    • hf.co/sarvamai/sarvam-30b-gguf:Q4_K_MLinux · x86-64 · no GPU · 32 GiB · unavailable · 2 grams

      the ref, or the quant asked of it, is not published upstream

    • hf.co/unsloth/gemma-3-27b-it-GGUF:IQ3_XXSLinux · x86-64 · no GPU · 32 GiB · unavailable · 2 grams

      the ref, or the quant asked of it, is not published upstream

    • hf.co/unsloth/gemma-4-26B-A4B-it-GGUF:Q4_K_MLinux · x86-64 · no GPU · 32 GiB · unavailable · 2 grams

      the ref, or the quant asked of it, is not published upstream

    • hf.co/unsloth/Qwen3.6-35B-A3B-GGUF:IQ1_MLinux · x86-64 · no GPU · 32 GiB · unavailable · 2 grams

      the ref, or the quant asked of it, is not published upstream

    • hf.co/unsloth/Qwen3.6-35B-A3B-GGUF:Q4_K_MLinux · x86-64 · no GPU · 32 GiB · unavailable · 2 grams

      the ref, or the quant asked of it, is not published upstream

    • hf.co/unsloth/Qwen3.6-35B-A3B-GGUF:Q4_K_SLinux · x86-64 · no GPU · 32 GiB · unavailable · 2 grams

      the ref, or the quant asked of it, is not published upstream

    • hf.co/unsloth/Qwen3.6-35B-A3B-MTP-GGUF:Q4_K_SLinux · x86-64 · no GPU · 32 GiB · unavailable · 2 grams

      the ref, or the quant asked of it, is not published upstream

    • hf.co/unsloth/Qwen3.8-27B-GGUF:Q4_K_MLinux · x86-64 · no GPU · 32 GiB · unavailable · 2 grams

      the ref, or the quant asked of it, is not published upstream

    • ai4bharat/indicconformer_stt_hi_hybrid_ctc_rnnt_largeWindows · x86-64 · Intel GPU · 16 GiB · unloadable · 1 gram

      the engine has no loader for this architecture — a release changes that, never a clock

    • ai4bharat/indicconformer_stt_hi_hybrid_ctc_rnnt_largeLinux · x86-64 · NVIDIA · 64 GiB · unloadable · 1 gram

      the engine has no loader for this architecture — a release changes that, never a clock

    • ai4bharat/indicconformer_stt_hi_hybrid_ctc_rnnt_largeLinux · x86-64 · no GPU · 32 GiB · unloadable · 1 gram

      the engine has no loader for this architecture — a release changes that, never a clock

    • bharatgenai/Shrutam-2Windows · x86-64 · Intel GPU · 16 GiB · unloadable · 1 gram

      the engine has no loader for this architecture — a release changes that, never a clock

    • bharatgenai/Shrutam-2Windows · x86-64 · NVIDIA · 16 GiB · unloadable · 1 gram

      the engine has no loader for this architecture — a release changes that, never a clock

    • bharatgenai/Shrutam-2Linux · x86-64 · NVIDIA · 64 GiB · unloadable · 1 gram

      the engine has no loader for this architecture — a release changes that, never a clock

    • bharatgenai/Shrutam-2Linux · x86-64 · no GPU · 32 GiB · unloadable · 1 gram

      the engine has no loader for this architecture — a release changes that, never a clock

    • bharatgenai/sooktam2Windows · x86-64 · Intel GPU · 16 GiB · unavailable · 1 gram

      the ref, or the quant asked of it, is not published upstream

    • bharatgenai/sooktam2Windows · x86-64 · NVIDIA · 16 GiB · unloadable · 1 gram

      the engine has no loader for this architecture — a release changes that, never a clock

    • bharatgenai/sooktam2Linux · x86-64 · NVIDIA · 64 GiB · unloadable · 1 gram

      the engine has no loader for this architecture — a release changes that, never a clock

    • bharatgenai/sooktam2Linux · x86-64 · no GPU · 32 GiB · unloadable · 1 gram

      the engine has no loader for this architecture — a release changes that, never a clock

    • coqui/XTTS-v2Windows · x86-64 · NVIDIA · 16 GiB · unavailable · 1 gram

      the ref, or the quant asked of it, is not published upstream

    • coqui/XTTS-v2macOS · ARM64 · Apple GPU · 16 GiB · unavailable · 1 gram

      the ref, or the quant asked of it, is not published upstream

    • coqui/XTTS-v2Windows · x86-64 · AMD · 8 GiB · unavailable · 1 gram

      the ref, or the quant asked of it, is not published upstream

    refusals are signed facts about (artifact, class): gated, unloadable, or absent upstream. They are coverage of a kind -- the fleet learned it once -- and they sink the cell to the back of every gram of that class. A refusal lifts when its season closes and the engine that refused has moved on.

    what proving costs, per hardware class
    • Windows · x86-64 · Intel GPU · 16 GiB5.8 MiB/s median fetch
      • generation: 3 s median probe · 1 sample
      • ollama: 25 min 34 s median probe · 4 samples
      • vision-task: 10 min 3 s median probe · 1 sample
      • speech-hf: 38 s median probe · 1 sample
      • diffusers: 1 min 22 s median probe · 1 sample
      • audiogen: 12 min median probe · 1 sample
      • translate-hf: 2 h 1 min median probe · 1 sample
    • Windows · x86-64 · NVIDIA · 16 GiB6.6 MiB/s median fetch
      • ollama: — median probe · 0 samples
      • ollama-host: 46 min 44 s median probe · 1 sample
      • generation: 0 s median probe · 1 sample
      • translate-hf: 13 min 9 s median probe · 1 sample
    • macOS · x86-64 · Apple GPU · 8 GiB3.5 MiB/s median fetch
      • Windows · x86-64 · AMD · 8 GiB590.1 MiB/s median fetch
        • ollama-host: 30 min 52 s median probe · 3 samples
        • generation: 1 s median probe · 1 sample
        • vision-task: 3 s median probe · 1 sample
        • speech-hf: 21 s median probe · 1 sample
        • ollama: — median probe · 0 samples
      • macOS · ARM64 · Apple GPU · 16 GiB20.2 MiB/s median fetch
        • ollama-host: 4 min 8 s median probe · 1 sample
        • ollama: — median probe · 0 samples
        • speech-hf: 8 min 5 s median probe · 1 sample
      • Windows · x86-64 · NVIDIA · 8 GiB7.3 MiB/s median fetch
        • Windows · x86-64 · no GPU · 8 GiB7.3 MiB/s median fetch
          • generation: 1 s median probe · 1 sample
          • ollama-host: 30 min 52 s median probe · 3 samples

          carried forward — no gram of this class reported in the latest aggregate

        • Windows · x86-64 · no GPU · 16 GiB5 MiB/s median fetch

            carried forward — no gram of this class reported in the latest aggregate

          • macOS · ARM64 · Apple GPU · 8 GiB9 MiB/s median fetch
            • ollama-host: 28 min 57 s median probe · 2 samples

            carried forward — no gram of this class reported in the latest aggregate

          • Linux · x86-64 · NVIDIA · 32 GiB105.9 MiB/s median fetch
            • speech-pod: 44 s median probe · 1 sample
            • generation: 2 min 45 s median probe · 1 sample
            • ollama-host: 1 h 27 min median probe · 1 sample
            • ollama: — median probe · 0 samples
            • vllm: — median probe · 0 samples
          • Linux · x86-64 · no GPU · 32 GiB73.3 MiB/s median fetch
            • speech-pod: 13 s median probe · 2 samples
            • ollama-host: 44 min 51 s median probe · 1 sample
            • ollama: — median probe · 0 samples
          • Linux · x86-64 · no GPU · 4 GiBfetch speed not measured
            • Linux · x86-64 · NVIDIA · 256 GiBfetch speed not measured
              • ollama: — median probe · 0 samples
            • Linux · x86-64 · NVIDIA · 64 GiB116.8 MiB/s median fetch
              • ollama: — median probe · 0 samples
              • vllm: — median probe · 0 samples
              • speech-pod: 2 s median probe · 1 sample
              • vision-task: 6 s median probe · 1 sample
            • ios · ARM64 · no GPU · 4 GiBfetch speed not measured

              cost is operational: medians of node-reported pull/probe clocks per hardware class and engine, with sample counts. It is not a measurement of any model and is never ranked.

              what the network actually used
              • hf.co/unsloth/gemma-4-E2B-it-GGUF:Q4_K_M512.1k tokens served
              • claude-haiku-4-5318.3k tokens served
              • gemma4:e2b282.3k tokens served
              • hf.co/Qwen/Qwen2.5-0.5B-Instruct-GGUF:q4_k_m237.5k tokens served
              • hf.co/Qwen/Qwen2.5-0.5B-Instruct-GGUF:Q4_K_M200.1k tokens served
              • hf.co/QuantFactory/OpenHathi-7B-Hi-v0.1-Base-GGUF:Q4_K_M155.7k tokens served
              • hf.co/unsloth/Qwen3-4B-GGUF:IQ3_XXS152.6k tokens served
              • hf.co/unsloth/Qwen3-4B-GGUF:IQ1_S127.6k tokens served
              • llama3.2:1b107.2k tokens served
              • hf.co/unsloth/Phi-4-mini-reasoning-GGUF:IQ1_S85.3k tokens served
              • hf.co/unsloth/GLM-4.7-Flash-GGUF:Q4_K_M84.3k tokens served
              • hf.co/unsloth/Qwen3.5-0.8B-GGUF:Q4_K_M83.7k tokens served

              demand is tokens produced per artifact per class, from signed throughput observations; it orders depth, never existence.

              Every number is a signed measurement or a count of them. grams is how many distinct nodes contributed: a verified floor, not a complete account. Lineage (publisher, attribution, base, cohort, rung) is from the catalog and is not a measurement.

              grams and classes are what THIS window heard from; coverage, refusals, cost and demand carry across windows, because a gram going to sleep does not unmake what it measured.

              The season file → All seasons →