Inference · /inferences/seasons/2026-W35

Season 2026-W35.

24 Aug 2026 to 31 Aug 2026. 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.

6grams contributed

28artifacts proved

2hardware classes

5publishers

cohorts this season

0 · Lead member, rung 00
1 · Lead member's ladder4
2 · Other vendor members4
3 · Derivatives4

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

hardware classes that signed

hardware not stated in the signed payload
Windows · x86-64 · no GPU · 16 GiB

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

publishers, by what the network proved of their builds

unsloth14 held probes · 5 grams · ollama 14
bartowski4 held probes · 2 grams · ollama 4
MaziyarPanahi4 held probes · 1 gram · ollama 4
mradermacher2 held probes · 2 grams · ollama 2
LanguaMan1 held probe · 1 gram · audiogen 1

Listed by held outcomes. Downloads, likes and popularity do not appear on this page, because nobody signs them.

cells the network has signed as unprovable

bharatgenai/sooktam2Windows · x86-64 · Intel GPU · 16 GiB · unavailable · 1 gram
the ref, or the quant asked of it, is not published upstream
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 · NVIDIA · 8 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
ResembleAI/chatterboxmacOS · ARM64 · Apple GPU · 16 GiB · unavailable · 1 gram
the ref, or the quant asked of it, is not published upstream
SWivid/F5-TTSWindows · x86-64 · NVIDIA · 16 GiB · unavailable · 1 gram
the ref, or the quant asked of it, is not published upstream
SWivid/F5-TTSmacOS · ARM64 · Apple GPU · 16 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 GiB5.2 MiB/s median fetch
- generation: 6 min 3 s median probe · 1 sample
Windows · x86-64 · NVIDIA · 16 GiB281.3 MiB/s median fetch
- ollama: 17 min 42 s median probe · 1 sample
- ollama-host: 46 min 44 s median probe · 1 sample
- generation: 0 s median probe · 1 sample
macOS · x86-64 · Apple GPU · 8 GiB906.1 MiB/s median fetch
carried forward — no gram of this class reported in the latest aggregate
Windows · x86-64 · AMD · 8 GiB894.6 MiB/s median fetch
- ollama-host: 30 min 52 s median probe · 3 samples
- generation: 1 s median probe · 1 sample
macOS · ARM64 · Apple GPU · 16 GiB409.8 MiB/s median fetch
- ollama-host: 4 min 8 s median probe · 1 sample
Windows · x86-64 · NVIDIA · 8 GiB13.8 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: 30 min 18 s median probe · 3 samples

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

gemma4:e2b213.2k tokens served
hf.co/unsloth/Qwen3-4B-GGUF:IQ3_XXS124.5k tokens served
claude-haiku-4-5100.8k tokens served
hf.co/unsloth/Qwen3-4B-GGUF:IQ1_S70.2k tokens served
hf.co/unsloth/gemma-4-E2B-it-GGUF:Q4_K_M58.7k tokens served
hf.co/Qwen/Qwen2.5-0.5B-Instruct-GGUF:Q4_K_M42.0k tokens served
llama3.2:1b40.4k tokens served
hf.co/lmstudio-community/Phi-4-mini-reasoning-GGUF:Q8_037.5k tokens served
hf.co/unsloth/Phi-4-mini-reasoning-GGUF:IQ1_S35.0k tokens served
hf.co/Qwen/Qwen3-4B-GGUF:Q5_023.8k tokens served
hf.co/Qwen/Qwen2.5-0.5B-Instruct-GGUF:q4_k_m18.5k tokens served
hf.co/lmstudio-community/Phi-4-mini-reasoning-GGUF:Q4_K_M13.5k tokens served

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

Cohort 1 · Lead member's ladder · 4

hf.co/MaziyarPanahi/Phi-4-mini-instruct-GGUF:Q3_K_SollamaQ3_K_Srung 2
MaziyarPanahi · quantization of microsoft/Phi-4-mini-instruct
- hardware not stated in the signed payload — 1 gram · 27.89 tok/s best · 4,096 ctx proved · 6 capabilities held
held on hardware not stated in the signed payload: code.write, lang.en, output.long, output.schema, research.synthesize, structured_output
served 163 tokens across the network
hf.co/unsloth/Qwen3-4B-GGUF:IQ3_XXSollamaIQ3_XXSrung 2
unsloth · quantization of Qwen/Qwen3-4B
- hardware not stated in the signed payload — 1 gram · 0.22 tok/s best · 4,096 ctx proved · 17 capabilities held
held on hardware not stated in the signed payload: code.reason, code.write, faithful_tool_result, instruct.follow, lang.en, lang.ta, output.long, output.schema, reason.channel, research.synthesize, structured_output, thinking, tools, tools.call, tools.faithful, tools.loop_ends, tools_loop_terminated
served 124.5k tokens across the network
hf.co/MaziyarPanahi/Llama-3.2-3B-Instruct-GGUF:IQ1_SollamaIQ1_Srung 5
MaziyarPanahi · quantization of meta-llama/Llama-3.2-3B-Instruct
- hardware not stated in the signed payload — 1 gram · 45.74 tok/s best · 3 capabilities held
held on hardware not stated in the signed payload: output.long, output.schema, structured_output
served 58 tokens across the network
hf.co/unsloth/Qwen3-4B-GGUF:IQ1_SollamaIQ1_Srung 5
unsloth · quantization of Qwen/Qwen3-4B
- hardware not stated in the signed payload — 1 gram · 24.9 tok/s best
served 70.2k tokens across the network

Cohort 2 · Other vendor members · 4

hf.co/MaziyarPanahi/Llama-3.2-1B-Instruct-GGUF:IQ3_XSollamaIQ3_XSrung 2
MaziyarPanahi · quantization of meta-llama/Llama-3.2-1B-Instruct
- hardware not stated in the signed payload — 1 gram · 107.53 tok/s best · 4,096 ctx proved · 13 capabilities held
held on hardware not stated in the signed payload: code.write, faithful_tool_result, instruct.follow, lang.en, lang.hi, output.long, output.schema, structured_output, tools, tools.call, tools.faithful, tools.loop_ends, tools_loop_terminated
served 1.4k tokens across the network
hf.co/bartowski/Llama-3.2-1B-Instruct-GGUF:F16ollamaF16rung 4
bartowski · quantization of meta-llama/Llama-3.2-1B-Instruct
- hardware not stated in the signed payload — 1 gram · 6.72 tok/s best · 14 capabilities held
held on hardware not stated in the signed payload: code.write, faithful_tool_result, instruct.follow, lang.en, lang.hi, lang.nl, lang.te, output.schema, structured_output, tools, tools.call, tools.faithful, tools.loop_ends, tools_loop_terminated
served 658 tokens across the network
hf.co/MaziyarPanahi/Llama-3.2-1B-Instruct-GGUF:IQ1_SollamaIQ1_Srung 5
MaziyarPanahi · quantization of meta-llama/Llama-3.2-1B-Instruct
- hardware not stated in the signed payload — 1 gram · 105.02 tok/s best
served 958 tokens across the network
hf.co/unsloth/Phi-4-mini-reasoning-GGUF:IQ1_SollamaIQ1_Srung 5
unsloth · quantization of microsoft/Phi-4-mini-reasoning
- hardware not stated in the signed payload — 2 grams · 26.35 tok/s best · 1 capabilities held
held on hardware not stated in the signed payload: lang.en
served 35.0k tokens across the network

Cohort 3 · Derivatives · 4

hf.co/bartowski/Llama-3.2-3B-Instruct-uncensored-GGUF:IQ3_XSollamaIQ3_XSrung 2
bartowski · derivative of chuanli11/Llama-3.2-3B-Instruct-uncensored
- hardware not stated in the signed payload — 1 gram · 46.8 tok/s best · 8,192 ctx proved · 10 capabilities held
held on hardware not stated in the signed payload: code.write, instruct.follow, lang.en, output.long, output.schema, structured_output, tools, tools.call, tools.loop_ends, tools_loop_terminated
served 45 tokens across the network
hf.co/bartowski/microsoft_Phi-4-mini-instruct-GGUF:IQ2_MollamaIQ2_Mrung 5
bartowski · derivative of microsoft/Phi-4-mini-instruct
- hardware not stated in the signed payload — 1 gram · 37.31 tok/s best · 4,096 ctx proved · 6 capabilities held
held on hardware not stated in the signed payload: code.write, instruct.follow, lang.en, output.long, output.schema, structured_output
served 58 tokens across the network
hf.co/mradermacher/Qwen3-VL-8B-Instruct-abliterated-GGUF:Q2_KollamaQ2_Krung 5
mradermacher · derivative of prithivMLmods/Qwen3-VL-8B-Instruct-abliterated-v1
- hardware not stated in the signed payload — 2 grams · 22.43 tok/s best · 4,096 ctx proved · 17 capabilities held
held on hardware not stated in the signed payload: code.write, instruct.follow, lang.bn, lang.en, lang.hi, lang.ta, output.long, output.schema, structured_output, tools, tools.call, tools.loop_ends, tools_loop_terminated, vision, vision.count, vision.read, vision.spatial
served 1.4k tokens across the network
LanguaMan/stable-audio-open-1-0-modelaudiogenrung 9
LanguaMan · derivative of a vendor base
- Windows · x86-64 · no GPU · 16 GiB — 1 gram

Measured, not in the catalog · 16

claude-haiku-4-5
unknown publisher
- hardware not stated in the signed payload — 1 gram · 4,096 ctx proved · 1 capabilities held
held on hardware not stated in the signed payload: tools
served 100.8k tokens across the network
claude-opus-5
unknown publisher
- hardware not stated in the signed payload — 1 gram · 1 capabilities held
held on hardware not stated in the signed payload: tools
served 5.3k tokens across the network
claude-sonnet-5
unknown publisher
- hardware not stated in the signed payload — 1 gram · 4,096 ctx proved · 1 capabilities held
held on hardware not stated in the signed payload: tools
hf.co/bartowski/Llama-3.2-1B-Instruct-GGUF:Q4_K_M
unknown publisher
- hardware not stated in the signed payload — 1 gram · 103.86 tok/s best · 8,192 ctx proved · 14 capabilities held
- Windows · x86-64 · no GPU · 16 GiB — 1 gram · 11.62 tok/s best · 4,096 ctx proved · 14 capabilities held
held on hardware not stated in the signed payload: code.write, faithful_tool_result, instruct.follow, lang.en, lang.ml, output.long, output.schema, research.synthesize, structured_output, tools, tools.call, tools.faithful, tools.loop_ends, tools_loop_terminated
served 5.0k tokens across the network
hf.co/bartowski/microsoft_Phi-4-mini-instruct-GGUF:Q3_K_L
unknown publisher
- hardware not stated in the signed payload — 1 gram · 32.06 tok/s best · 8,192 ctx proved · 9 capabilities held
held on hardware not stated in the signed payload: code.write, instruct.follow, lang.bn, lang.en, lang.hi, output.long, output.schema, research.synthesize, structured_output
served 16 tokens across the network
hf.co/unsloth/gemma-4-E2B-it-GGUF:IQ4_NL
unknown publisher
- hardware not stated in the signed payload — 1 gram · 54.91 tok/s best · 31 capabilities held
held on hardware not stated in the signed payload: code.reason, code.write, faithful_tool_result, instruct.follow, lang.bn, lang.en, lang.gu, lang.hi, lang.kn, lang.ml, lang.mr, lang.pa, lang.ta, lang.te, output.long, output.schema, reason.channel, research.synthesize, structured_output, thinking, tools, tools.call, tools.faithful, tools.loop_ends, tools.multi_hop, tools_loop_terminated, vision, vision.count, vision.multi_image, vision.read, vision.spatial
served 3.7k tokens across the network
hf.co/unsloth/gemma-4-E2B-it-GGUF:Q3_K_M
unknown publisher
- hardware not stated in the signed payload — 1 gram · 38.36 tok/s best · 8,192 ctx proved · 31 capabilities held
held on hardware not stated in the signed payload: code.reason, code.write, faithful_tool_result, instruct.follow, lang.bn, lang.en, lang.gu, lang.hi, lang.kn, lang.ml, lang.mr, lang.pa, lang.ta, lang.te, output.long, output.schema, reason.channel, research.synthesize, structured_output, thinking, tools, tools.call, tools.faithful, tools.loop_ends, tools.multi_hop, tools_loop_terminated, vision, vision.count, vision.multi_image, vision.read, vision.spatial
served 2.2k tokens across the network
hf.co/unsloth/gemma-4-E2B-it-GGUF:Q4_0
unknown publisher
- Windows · x86-64 · no GPU · 16 GiB — 1 gram · 9.1 tok/s best · 4,096 ctx proved · 31 capabilities held
held on Windows · x86-64 · no GPU · 16 GiB: code.reason, code.write, faithful_tool_result, instruct.follow, lang.bn, lang.en, lang.gu, lang.hi, lang.kn, lang.ml, lang.mr, lang.pa, lang.ta, lang.te, output.long, output.schema, reason.channel, research.synthesize, structured_output, thinking, tools, tools.call, tools.faithful, tools.loop_ends, tools.multi_hop, tools_loop_terminated, vision, vision.count, vision.multi_image, vision.read, vision.spatial
served 4.4k tokens across the network
hf.co/unsloth/gemma-4-E4B-it-GGUF:Q3_K_M
unknown publisher
- hardware not stated in the signed payload — 1 gram · 20.35 tok/s best · 4,096 ctx proved · 30 capabilities held
held on hardware not stated in the signed payload: code.reason, code.write, faithful_tool_result, instruct.follow, lang.bn, lang.en, lang.gu, lang.hi, lang.kn, lang.ml, lang.pa, lang.ta, lang.te, output.long, output.schema, reason.channel, research.synthesize, structured_output, thinking, tools, tools.call, tools.faithful, tools.loop_ends, tools.multi_hop, tools_loop_terminated, vision, vision.count, vision.multi_image, vision.read, vision.spatial
served 1.2k tokens across the network
hf.co/unsloth/Phi-4-mini-reasoning-GGUF:Q2_K
unknown publisher
- hardware not stated in the signed payload — 1 gram · 37.71 tok/s best · 8 capabilities held
- Windows · x86-64 · no GPU · 16 GiB — 1 gram · 9.83 tok/s best · 5 capabilities held
held on hardware not stated in the signed payload: code.reason, lang.en, output.long, output.schema, reason.channel, research.synthesize, structured_output, thinking
served 2.1k tokens across the network
hf.co/unsloth/Phi-4-mini-reasoning-GGUF:Q2_K_L
unknown publisher
- hardware not stated in the signed payload — 2 grams
served 3.9k tokens across the network
hf.co/unsloth/Phi-4-mini-reasoning-GGUF:Q3_K_M
unknown publisher
- hardware not stated in the signed payload — 2 grams · 31.94 tok/s best · 8 capabilities held
held on hardware not stated in the signed payload: code.reason, lang.en, output.long, output.schema, reason.channel, research.synthesize, structured_output, thinking
served 1.2k tokens across the network
hf.co/unsloth/Phi-4-mini-reasoning-GGUF:Q4_K_M
unknown publisher
- Windows · x86-64 · no GPU · 16 GiB — 1 gram · 6.5 tok/s best · 6 capabilities held
held on Windows · x86-64 · no GPU · 16 GiB: code.reason, output.schema, reason.channel, research.synthesize, structured_output, thinking
served 412 tokens across the network
hf.co/unsloth/Qwen3-VL-4B-Instruct-GGUF:Q4_0
unknown publisher
- hardware not stated in the signed payload — 3 grams · 40.87 tok/s best · 4,096 ctx proved · 21 capabilities held
held on hardware not stated in the signed payload: code.write, faithful_tool_result, instruct.follow, lang.en, lang.hi, lang.kn, lang.ml, lang.ta, lang.te, output.long, output.schema, structured_output, tools, tools.call, tools.faithful, tools.loop_ends, tools_loop_terminated, vision, vision.count, vision.read, vision.spatial
served 580 tokens across the network
hf.co/unsloth/Qwen3-VL-4B-Instruct-GGUF:Q4_1
unknown publisher
- hardware not stated in the signed payload — 1 gram · 37.02 tok/s best · 4,096 ctx proved · 20 capabilities held
held on hardware not stated in the signed payload: code.write, faithful_tool_result, instruct.follow, lang.en, lang.hi, lang.ml, lang.pa, lang.te, output.long, output.schema, structured_output, tools, tools.call, tools.faithful, tools.loop_ends, tools_loop_terminated, vision, vision.count, vision.read, vision.spatial
served 378 tokens across the network
llama3.2:1b
unknown publisher
- hardware not stated in the signed payload — 1 gram · 13.09 tok/s best · 10 capabilities held
held on hardware not stated in the signed payload: code.write, instruct.follow, lang.en, output.long, output.schema, structured_output, tools, tools.call, tools.loop_ends, tools_loop_terminated
served 40.4k tokens across the network

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 →