Showing 50 of 50 submissions

0 selected β€” tick 2–4 rows to compare
Community submissions grouped by VRAM tier. Each row is one setup; column headers sort the table, and rows open a detail panel.
Actions
Under 16 GB
Novasty ● RTX 3060 12GB NVIDIA Other HauhauCS/Qwen3.6-35B-A3B-Uncensored-HauhauCS-Aggressive GGUF Q4_K_M RAG 14.8 196608 Default 170w Hybrid (VRAM+RAM)
16–23 GB
Burhan ● 1x RTX-4000 Ada SFF (20 GB) NVIDIA Ollama Qwen3.6-27b-q4_KM GGUF Q4_K_M Document Q&A, Deep Research, Brainstorming, Coding ~20-30 t/s 52000 q8_0 75 W Hybrid (VRAM+RAM)
Burhan ● 1x RTX-4000 Ada SFF (20 GB) NVIDIA Ollama LFM2.5-thinking:1.2b-q8_0 GGUF Q8_0 Q&A, Brainstorming, Complex Tasks ~35 t/s 64000 q8_0 75 W Hybrid (VRAM+RAM)
Burhan ● 1x RTX-4000 Ada SFF (20 GB) NVIDIA Ollama Gemma4:26b-a4b-it-mtp-q4_K_M GGUF Q4_K_M Document Q&A, Deep Research, Creative Writing ~25-30 t/s 64000 q8_0 75 W Hybrid (VRAM+RAM)
Burhan ● 1x RTX-4000 Ada SFF (20 GB) NVIDIA Ollama Qwen3.5:9b-q8_0 GGUF Q8_0 Document Q&A, Deep Research, Creative Writing, Brainstorming, Planning ~30 t/s 128000 q8_0 75 W Hybrid (VRAM+RAM)
s1ugh34d ● Radeon VII 16GB HBM2 (gfx906), 32-core EPYC host, 67GB RAM AMD llama-swap samuelcardillo/Carnice-Qwen3.6-MoE-35B-A3B (I-Nano) GGUF I-Nano LLM assited coding / technical documentation 54.4 65536 tokens native (NIAH hit YES at 65k) quantized KV (observed near-flat VRAM growth w/ context) ~150 W peak, 99% GPU util
s1ugh34d ● Radeon VII 16GB HBM2 (gfx906) AMD llama-swap deadbydawn101/gemma-4-E4B-Agentic-Opus-Reasoning-GeminiCLI-GGUF GGUF Q4_K_M LLM assited coding / technical documentation / large repo generation 86.7 49152 tokens (NIAH hit YES at 49k; 65k not probed) quantized KV (VRAM ~flat: 4618MB @49k vs 4619MB @4k) ~214 W peak, 99% GPU util
s1ugh34d ● Radeon VII 16GB HBM2 (gfx906) AMD llama-swap unsloth/Qwen3.8-27B-GGUF (UD-Q2_K_XL) GGUF UD-Q2_K_XL LLM assited coding / technical documentation 13.9 49152 tokens (NIAH hit YES at 49k) quantized KV (VRAM 13770β†’13798MB across 8kβ†’32k, +28MB) ~233 W peak, 100% GPU util
s1ugh34d ● Radeon VII 16GB HBM2 (gfx906) AMD llama-swap MaziyarPanahi/DeepSeek-R1-0528-Qwen3-8B-GGUF GGUF Q4_K_M LLM assited coding / technical documentation 84.4 16384 tokens (native server cap) quantized KV (7327β†’7299 MB, flat) ~268 W peak, 93% GPU util
s1ugh34d ● Radeon VII 16GB HBM2 (gfx906) AMD llama-swap bartowski/Qwen_Qwen3.5-4B-GGUF GGUF Q4_K_M LLM assited coding / technical documentation 96.4 65536 tokens (native cap; NIAH hit YES; 96k rejected, n_ctx=65536 enforced) quantized KV (VRAM 5312MB @4k β†’ 5312MB @65k, +0) ~225 W peak, 99% GPU util
Yehya_Karout ● 1x RTX 5080 16GB + 64GB DDR5-6400, Core Ultra 9 285K NVIDIA Strata Qwen3.8-Flash-Next (GSQ-RCO GGUF) IQ3_XXS, int8 KV agentic coding (opencode, pi), full 262K context ~170 tok/s tg / ~3,000 pp at 4K, ~150 tg / ~3,950 pp at 180K, 152 tg / 3,800 pp at 250K. Real opencode agent session: ~165-170 tok/s avg
f41n ● 1xRX 6800 XT 16GB + Ryzen 7 7700X + 32GB DDR5-5200 llama.cpp gpt-oss 20B Q8_0 coding, agents 131 t/s generation, 2762 t/s prompt
f41n ● 1x RX 6800 XT 16GB + Ryzen 7 7700X + 32GB DDR5-5200 llama.cpp Nemotron3 Nano 4B Q4_K_M quick tasks, small helper model 123 t/s generation, 2412 t/s prompt
f41n ● 1x RX 6800 XT 16GB + Ryzen 7 7700X + 32GB DDR5-5200 llama.cpp Gemma 4 26B-A4B UD-IQ4_XS daily chat, general 73 t/s generation, 2278 t/s prompt
f41n ● 1x RX 6800 XT 16GB + Ryzen 7 7700X + 32GB DDR5-5200 llama.cpp Qwen3.8-27B 3-bit GSQ-RCO (mixed IQ3_S etc.) + MTP head coding, agents \~51 t/s generation
f41n ● 1x RX 6800 XT 16GB + Ryzen 7 7700X + 32GB DDR5-5200 llama.cpp Qwen3.8-27B 3-bit GSQ-RCO (mixed IQ3_S etc.) coding, agents 27 t/s generation, 441 t/s prompt
f41n ● 1x RX 6800 XT 16GB + Ryzen 7 7700X + 32GB DDR5-5200 llama.cpp Qwen3.8-Flash-Next (125B MoE) GSQ-RCO Q2_0 (37.6 GB shard 1 + 29 GB n-gram table shard 2) coding, agents \~12 t/s generation (warm), \~12 t/s prompt, \~5 t/s first cold prompt
f41n ● 1x Radeon Instinct MI25 16GB HBM2 @ 170W (WX 9100 BIOS) + Ryzen 5 3600 + 32GB DDR4-3200 AMD llama.cpp gpt-oss 20B Q8_0 coding, agents 89 t/s generation, 858 t/s prompt
f41n ● 1x Radeon Instinct MI25 16GB HBM2 @ 170W (WX 9100 BIOS) + Ryzen 5 3600 + 32GB DDR4-3200 AMD llama.cpp Nemotron3 Nano 4B Q4_K_M quick tasks, small helper model 77 t/s generation, 783 t/s prompt
f41n ● 1x Radeon Instinct MI25 16GB HBM2 @ 170W (WX 9100 BIOS) + Ryzen 5 3600 + 32GB DDR4-3200 AMD llama.cpp Qwen3.6 35B-A3B IQ3_S (Unsloth UD) coding, agents 68 t/s generation, 586 t/s prompt
f41n ● 1x Radeon Instinct MI25 16GB HBM2 @ 170W (WX 9100 BIOS) + Ryzen 5 3600 + 32GB DDR4-3200 AMD llama.cpp Gemma 4 26B-A4B UD-IQ4_XS daily chat, general 57 t/s generation, 678 t/s prompt
f41n ● 1x Radeon Instinct MI25 16GB HBM2 @ 170W (WX 9100 BIOS) + Ryzen 5 3600 + 32GB DDR4-3200 AMD llama.cpp Qwen3.8-27B 3-bit GSQ-RCO (mixed IQ3_S etc.) + MTP head coding, agents \~23 t/s generation
f41n ● 1x Radeon Instinct MI25 16GB HBM2 @ 170W (WX 9100 BIOS) + Ryzen 5 3600 + 32GB DDR4-3200 AMD llama.cpp Qwen3.8-27B 3-bit GSQ-RCO (mixed IQ3_S etc.) coding, agents 17 t/s generation, 134 t/s prompt
f41n ● 1x Radeon Instinct MI25 16GB HBM2 @ 170W (WX 9100 BIOS) + Ryzen 5 3600 + 32GB DDR4-3200 AMD llama.cpp Qwen3.8-Flash-Next (125B MoE) GSQ-RCO Q2_0 (37.6 GB shard 1 + 29 GB n-gram table shard 2) coding, agents \~17 t/s generation (warm, 15.5-18), \~14 t/s prompt, \~10 t/s first cold prompt
f41n ● 1x RX 6800 XT 16GB + Ryzen 7 7700X + 32GB DDR5-5200 llama.cpp LFM2.5 8B-A1B Q4_K_M quick tasks, small helper model 221 t/s generation, 4718 t/s prompt
f41n ● 1x RX 6800 XT 16GB + Ryzen 7 7700X + 32GB DDR5-5200 llama.cpp LFM2.5 8B-A1B Q4_K_M + DSpark draft (Q4_K_M) quick tasks, small helper model 240 t/s generation
24–31 GB
Kif ● 1x RTX 4090 24GB, Ryzen 9 5900X, 128GB DDR4-3600 CL18 (2x CMK64GX4M2D3600C18, 4x32GB double-rank, holding 3600 MT/s XMP), Gigabyte Aorus X570 Master, Samsung 990 Pro NVMe (PCIe 4.0) NVIDIA Strata Qwen/Qwen3.8-Flash-Next (GGUF IQ3_S) GGUF IQ3_S (ISTA-DASLab GSQ-RCO) Agentic coding / agent harness via OpenAI-compatible API 128 131072 configured; >=106,493 tokens observed in live agent sessions, no OOM, no coherence collapse int8 (engine-native quantized KV) 450W stock cap, 12W idle at P8; peak under load not measured
Kif ● 1x RTX 4090 24GB, Ryzen 9 5900X, 128GB DDR4-3600 CL18 (4x32GB double-rank), Gigabyte Aorus X570 Master, Samsung 990 Pro NVMe NVIDIA Strata Qwen/Qwen3.8-Flash-Next (GGUF IQ3_S) GGUF IQ3_S (ISTA-DASLab GSQ-RCO) Agentic coding / agent harness via OpenAI-compatible API (+ vision via mmproj) 107 262144 configured; 228,553 tokens observed in live sessions (65 s cold prefill), no OOM, no truncation, no coherence collapse int8 450W stock cap, 12W idle at P8
Kif ● 1x RTX 4090 24GB, Ryzen 9 5900X, 128GB DDR4-3600 CL18 (4x32GB double-rank), Gigabyte Aorus X570 Master, Samsung 990 Pro NVMe NVIDIA Strata Qwen/Qwen3.8-Flash-Next (GGUF IQ3_S) GGUF IQ3_S (ISTA-DASLab GSQ-RCO) Agentic coding, 2 parallel streams via OpenAI-compatible API 39.9 262144 per slot x 2 slots; 2x41k-token parallel fresh streams measured with KV in host RAM, plus a single 255k-token request (74 s cold prefill) k8v4 + --kv-resident 32768 (32,768 positions per layer in VRAM, rest in host RAM; injected via config user_args since the engine declines it under WSL2) 450W stock cap, 12W idle at P8
48–95 GB
digitalscream ● 2 x R9700 AMD vLLM d0xin/Swift-Qwen3.8-27B-Uncensored-FP8 FP8 Coding, Research (OpenCode) 120t/s code, 70t/s chat ~240k FP16
manoreken2 ● RTX 5090 + IceLake Xeon NVIDIA llama.cpp nemotron-3-super:120b
manoreken2 ● RTX 5090 + IceLake Xeon NVIDIA llama.cpp GPT-OSS:120b
PawingUnplug ● 2 x R9700 AMD llama.cpp bartowski/Kwaipilot_KAT-Coder-V2.5-Dev-Q5_K_M.gguf, Qwen3-Embedding-8B-Q8_0.gguf, Qwen3.5-9B-The-Defiant-Fable-Uncnr-Heretic-NEO-MAX-MTP-Q8_0.gguf Coding Judge, Research (OpenCode and own harness), Embeddings ~50t/s chat ~262k FP16 ~600w
PawingUnplug ● 2 x w7900 AMD llama.cpp Qwen3.8-Flash-Next-Q4_K_M Coding ~35t/s chat ~262k FP16 ~750w
digitalscream ● 2 x R9700 32GB, Threadripper 3960X, 128GB DDR4 AMD Radiance Swift 1.5 Flash Next IQ4 weights, FP8 PLE General purpose / agentic coding / agent harness via OpenAI-compatible API 200 262,144 configured; >=200,000 tokens observed in live agent sessions, no OOM, no coherence collapse FP8 500-600W during inference, GPUs at 330W cap each
mattlach ● 1x AMD Instinct MI210 passed through to VM w. 32GB of RAM AMD llama.cpp Gemma4:26B Q8_0 Daily chat, basic scripting, summarization Pre-fill: ~2,200 tok/s. Eval ~75-80 tok/s
mattlach ● 1x AMD Instinct MI210 passed through to VM w. 32GB of RAM AMD llama.cpp Gemma4:12B Q8_0 Model dedicated to context compaction for other models. Pre-fill: ~1,300 tok/s. Eval ~40-45 tok/s
mattlach ● 1x AMD Instinct MI210 passed through to VM w. 32GB of RAM AMD llama.cpp Gemma4:E4B Q4_K_M Model dedicated to small front end tasks (title generation, tags, etc.) Pre-fill: ~2,500 - 3,000 tok/s. Eval ~85-90 tok/s
mattlach ● 1x AMD Instinct MI210 passed through to VM w. 32GB of RAM AMD llama.cpp Jina Embeddings V3 F16 Model dedicated for embeddings
96+ GB
manoreken2 ● RTX 5090 + IceLake Xeon NVIDIA llama.cpp DeepSeek-R1:671b-0528_q8_0 GGUF Q8_0 1 tokens/sec 8192 default
manoreken2 ● RTX 5090 + IceLake Xeon NVIDIA llama.cpp DeepSeek-V3.1 Terminus:671b_q8_0 GGUF Q8_0 1 tokens/sec
manoreken2 ● RTX 5090 + IceLake Xeon NVIDIA llama.cpp DeepSeek-V3.2:671b_UD-Q8_K_XL 1 tokens/sec
manoreken2 ● RTX 5090 + IceLake Xeon NVIDIA llama.cpp GLM-5.2-UD-Q6_K_XL 1 tokens/sec
manoreken2 ● RTX 5090 + IceLake Xeon NVIDIA llama.cpp Kimi-K2.6-UD-Q8_K_XL 1 tokens/sec
manoreken2 ● RTX 5090 + IceLake Xeon NVIDIA llama.cpp Qwen3:235b-a22b-thinking-2507-q4
manoreken2 ● 1x RTX5090 (32 GB) + IceLake Xeon CPU (1024 GB) NVIDIA llama.cpp DeepSeek-V4-Pro-0813-UD_K_XL.gguf translation of government journal from 481 BCE China 16384
SgtAwesomesauce ● 4x AMD v620 AMD vLLM qwen3.8-flash-next W4A16 Agentic coding, chatbot tomfoolery 68-92 245K FP16/BF16 ~900w, full system power from wall
manoreken2 ● Intel Arc B60 24GB x4 Intel llama.cpp gemma-4-31B-it-BF16.gguf BF16 CtoJ translation of Chinese history θ³‡ζ²»ι€šι‘‘ 2.5
Unclassified (no VRAM tier)
manoreken2 ● Colibri DeepSeek-V4.1-Flash translation of government journal from 481 BCE China 12288
mattlach ● AMD Epyc 7543 (Milan, 32C/64T), LXC container with 128GB of RAM (no GPU) AMD Ollama Poolside Laguna S2.1 (118B) Q4_K_M Coding, and long form reasoning beyond what Gemma4:26B is capable of Pre-fill 60-62 tok/s. Evaluation 12-14 tok/s

The data at a glance

Three views of the 50 community submissions. The charts are generated from the data on every update β€” new submissions appear automatically.

VRAM distribution

VRAM distribution: 50 submissions across 6 tiers Bar chart of community submissions per VRAM tier. Under 16 GB: 1; 16–23 GB: 25; 24–31 GB: 3; 32–47 GB: 0; 48–95 GB: 10; 96+ GB: 9 0 10 20 30 Under 16 GB: 1 submission 1 Under 16 GB 16–23 GB: 25 submissions 25 16–23 GB 24–31 GB: 3 submissions 3 24–31 GB 32–47 GB: 0 submissions 0 32–47 GB 48–95 GB: 10 submissions 10 48–95 GB 96+ GB: 9 submissions 9 96+ GB VRAM tier

Backend usage

Backend usage: 50 submissions across 8 backends Donut chart of submissions per inference backend. llama.cpp: 31 (62%); llama-swap: 5 (10%); Ollama: 5 (10%); Strata: 4 (8%); vLLM: 2 (4%); Colibri: 1 (2%); Other: 1 (2%); Radiance: 1 (2%) llama.cpp: 31 submissions (62%) llama-swap: 5 submissions (10%) Ollama: 5 submissions (10%) Strata: 4 submissions (8%) vLLM: 2 submissions (4%) Colibri: 1 submission (2%) Other: 1 submission (2%) Radiance: 1 submission (2%) 50 submissions
  • llama.cpp 31
  • llama-swap 5
  • Ollama 5
  • Strata 4
  • vLLM 2
  • Colibri 1
  • Other 1
  • Radiance 1

Model size vs VRAM

Model size vs VRAM: 37 submissions plotted Scatter plot of total parameter count (x-axis, billions) against VRAM footprint (y-axis, GB). Swift-Qwen3.8-27B-Uncensored-FP8: 27B / 96GB; Qwen3.6-27b-q4_KM: 27B / 20GB; LFM2.5-thinking:1.2b-q8_0: 1.2B / 20GB; Gemma4:26b-a4b-it-mtp-q4_K_M: 26B / 20GB; Qwen3.5:9b-q8_0: 9B / 20GB; Carnice-Qwen3.6-MoE-35B-A3B: 35B / 12.7GB; gemma-4-E4B-Agentic-Opus-Reasoning-GeminiCLI-GGUF: 4B / 5GB; Qwen3.8-27B-GGUF: 27B / 13.8GB; DeepSeek-R1-0528-Qwen3-8B-GGUF: 8B / 7.4GB; Qwen_Qwen3.5-4B-GGUF: 4B / 16GB; DeepSeek-R1:671b-0528_q8_0: 671B / 1024GB; DeepSeek-V3.1 Terminus:671b_q8_0: 671B / 999GB; DeepSeek-V3.2:671b_UD-Q8_K_XL: 671B / 999GB; Qwen3:235b-a22b-thinking-2507-q4: 235B / 999GB; nemotron-3-super:120b: 120B / 96GB; GPT-OSS:120b: 120B / 96GB; gemma-4-31B-it-BF16.gguf: 31B / 24GB; Qwen3.6-35B-A3B-Uncensored-HauhauCS-Aggressive: 35B / 12GB; gpt-oss 20B: 20B / 16GB; Nemotron3 Nano 4B: 4B / 16GB; Gemma 4 26B-A4B: 26B / 16GB; Qwen3.8-27B: 27B / 16GB; Qwen3.8-27B: 27B / 16GB; Qwen3.8-Flash-Next: 125B / 16GB; gpt-oss 20B: 20B / 16GB; Nemotron3 Nano 4B: 4B / 16GB; Qwen3.6 35B-A3B: 35B / 16GB; Gemma 4 26B-A4B: 26B / 16GB; Qwen3.8-27B: 27B / 16GB; Qwen3.8-27B: 27B / 16GB; Qwen3.8-Flash-Next: 125B / 16GB; Gemma4:26B: 26B / 512GB; Gemma4:12B: 12B / 32GB; Gemma4:E4B: 4B / 32GB; Poolside Laguna S2.1: 118B / 128GB; LFM2.5 8B-A1B: 8B / 16GB; LFM2.5 8B-A1B: 8B / 16GB 0 200 400 600 800 0 500 1000 1500 LFM2.5-thinking:1.2b-q8_0: 1.2B params, ~20GB VRAM gemma-4-E4B-Agentic-Opus-Reasoning-GeminiCLI-GGUF: 4B params, ~5GB VRAM Qwen_Qwen3.5-4B-GGUF: 4B params, ~16GB VRAM Nemotron3 Nano 4B: 4B params, ~16GB VRAM Nemotron3 Nano 4B: 4B params, ~16GB VRAM Gemma4:E4B: 4B params, ~32GB VRAM DeepSeek-R1-0528-Qwen3-8B-GGUF: 8B params, ~7.4GB VRAM LFM2.5 8B-A1B: 8B params, ~16GB VRAM LFM2.5 8B-A1B: 8B params, ~16GB VRAM Qwen3.5:9b-q8_0: 9B params, ~20GB VRAM Gemma4:12B: 12B params, ~32GB VRAM gpt-oss 20B: 20B params, ~16GB VRAM gpt-oss 20B: 20B params, ~16GB VRAM Gemma4:26b-a4b-it-mtp-q4_K_M: 26B params, ~20GB VRAM Gemma 4 26B-A4B: 26B params, ~16GB VRAM Gemma 4 26B-A4B: 26B params, ~16GB VRAM Gemma4:26B: 26B params, ~512GB VRAM Swift-Qwen3.8-27B-Uncensored-FP8: 27B params, ~96GB VRAM Qwen3.6-27b-q4_KM: 27B params, ~20GB VRAM Qwen3.8-27B-GGUF: 27B params, ~13.8GB VRAM Qwen3.8-27B: 27B params, ~16GB VRAM Qwen3.8-27B: 27B params, ~16GB VRAM Qwen3.8-27B: 27B params, ~16GB VRAM Qwen3.8-27B: 27B params, ~16GB VRAM gemma-4-31B-it-BF16.gguf: 31B params, ~24GB VRAM Carnice-Qwen3.6-MoE-35B-A3B: 35B params, ~12.7GB VRAM Qwen3.6-35B-A3B-Uncensored-HauhauCS-Aggressive: 35B params, ~12GB VRAM Qwen3.6 35B-A3B: 35B params, ~16GB VRAM Poolside Laguna S2.1: 118B params, ~128GB VRAM nemotron-3-super:120b: 120B params, ~96GB VRAM GPT-OSS:120b: 120B params, ~96GB VRAM Qwen3.8-Flash-Next: 125B params, ~16GB VRAM Qwen3.8-Flash-Next: 125B params, ~16GB VRAM Qwen3:235b-a22b-thinking-2507-q4: 235B params, ~999GB VRAM DeepSeek-R1:671b-0528_q8_0: 671B params, ~1024GB VRAM DeepSeek-V3.1 Terminus:671b_q8_0: 671B params, ~999GB VRAM DeepSeek-V3.2:671b_UD-Q8_K_XL: 671B params, ~999GB VRAM Model size (billions of parameters) VRAM (GB)

Reading the matrix

Click any column header to sort (within each tier), or use the Sort dropdown for VRAM, model size, speed, or date across the whole table. Cells show the full value on hover. Speed is community-reported decode throughput; Context is the stable context tested; KV is the KV-cache precision; Mode is how the engine splits weights across VRAM/RAM/SSD. All data is community-reported β€” treat numbers as user-experienced, not lab-tested. Discussion lives in the forum thread β†’

The search box covers every field β€” model name, notes, hardware, backend, quant, use case. Backend and use case are multi-select: tick several, and they combine as any-of within the list and all-of across the other filters. Active filters appear as chips below the bar (each has an βœ•), and the count line tells you exactly what you're seeing.

Click any row (or its β“˜ button) to open the full submission β€” complete hardware, the submitter's notes, a launch command, and a deep link straight to the forum post. The πŸ”— Share button copies a link to this exact view β€” every filter, the sort, and the search included β€” so you can paste it in the forum and others land on the same table.

The Actions column: β“˜ opens the detail panel, tick the checkbox to add a row to a comparison, or hit πŸš€ to copy a launch command for that setup.