Machine
RX 9070 9070 Xt 9070 Gre · Ryzen 5 4500 #2d8cf276
89 runs of 28 models since Oct 4, 2026. Last run Oct 29, 2026.
89
runs
28
models
44
configs
20
configs run 2+ times
Hardware
- GPU
- AMD Radeon RX 9070/9070 XT/9070 GRE 16 GB
- CPU
- AMD Ryzen 5 4500 6-Core Processor 6 cores
- RAM
- 23 GB
- OS
- Ubuntu 24.04.5 LTS
- Tools
- llama.cpp 81 · Ollama 8
- Client
- 0.4.98
Models on this machine
One dot per model, at its best config. Up is better answers, right is faster. Time to finish counts every token the model wrote, thinking included, so a model that talks a lot ends up further left.
Best trade-off: quality 60+, and nothing else here is both faster and betterOther models · dot size = number of runs
Medians of each model's best config, 27 models.
Best config per model
sort
Runs with the same model, quant, tool, context, thinking and KV cache count as one config. Numbers are the median of its runs, and each dot is one run on the 0–100 quality scale. ★ marks the best config with at least 2 rated runs.
configrunsqualityeach runspeed
ISTA-DASLab/Qwen3.8-27B-GSQ-RCO-GGUF
5 runs · 3 configs
★ bestIQ3_XXS-mtpllama.cpp32k ctxthinking oneffort xhigh
×3 runs
85.672.5–87.0
varies
44.1 tok/s
42.5–44.5
ukisai/Swift-1.5-Qwen3.8-27B-GSQ-RCO-GGUF
5 runs · 3 configs
★ bestIQ3_S-mtpllama.cpp32k ctxthinking oneffort xhigh
×3 runs
85.283.3–85.3
steady
44.2 tok/s
43.9–44.4
prism-ml/Ternary-Bonsai-2-27B-gguf-PQ2_0_Vulkan
1 run · 1 config
peculiar-ragdoll/Tiel-Coder-35B-A3B-GGUF-MTP
2 runs · 2 configs
unsloth/Muse-Glimmer-30B-GGUF
5 runs · 2 configs
★ bestUD-IQ3_XXSllama.cpp32k ctxthinking on
×3 runs
84.582.8–88.4
steady
37.5 tok/s
37.5–37.6
pfeifferj/Ornith-1.5-35B-A3B-GSQ-RCO-GGUF
10 runs · 4 configs
★ best3bitllama.cpp32k ctxthinking default
×5 runs
84.462.2–86.7
2 outliers1 slow run
88.2 tok/s
56.8–104
ConwayResearch/Underdog-Saluki-27B-1.0
1 run · 1 config
prism-ml/Ternary-Bonsai-2-27B-gguf-PQ2_0
1 run · 1 config
IsValorum/Ornith-1.5-35B-A3B-APEX-I-NanoPlus-GGU
1 run · 1 config
IsValorum/Ornith-1.5-35B-A3B-APEX-I-NanoPlus-GGUF
3 runs · 1 config
★ bestllama.cpp32k ctxthinking default
×3 runs
82.181.4–84.3
steady
110 tok/s
93.5–110
Compare with another machine
Run history
Oct 29, 2026
20:59gemma-4-26B-A4B-it-qat-uncensored-heretic-UDmerge-GGUF:Q4_K_XXLOllama94.8 tok/s69.0Oct 10, 2026
16:25Qwen3-Coder-30B-A3B-Instruct-GGUF:UD-IQ3_XXS×2llama.cpp120 tok/s60.016:17Mellum2.1-12B-A2.5B-Thinking-GGUF:MXFP4_MOE×9llama.cpp162 tok/s68.113:08Muse-Glimmer-30B-GGUF:UD-IQ3_XXS×3llama.cpp37.5 tok/s84.512:06Muse-Glimmer-30B-GGUF:UD-IQ3_XXS×2llama.cpp37.7 tok/s81.309:29Underdog-Saluki-27B-1.0llama.cpp43.5 tok/s84.309:08Bielik-11B-v3.0-Instruct-GGUF:Q5_K_M×5llama.cpp57.8 tok/s55.1Oct 8, 2026
18:24Ornith-1.5-35B-A3B-GSQ-RCO-GGUF:3bitOllama102 tok/s83.918:17Ornith-1.5-35B-A3B-GSQ-RCO-GGUF:3bitOllama103 tok/s81.2Oct 7, 2026
20:46Ornith-1.5-35B-A3B-GSQ-RCO-GGUF:3bitOllama75.4 tok/s66.819:47Ornith-1.5-35B-A3B-GSQ-RCO-GGUF:3bitOllama75.1 tok/s79.6Oct 6, 2026
19:03Occamy-1.0-APEX-I-NanoPlus-GGUF×3llama.cpp90.1 tok/s69.918:43Occamy-1.0-APEX-I-NanoPlus-GGUFllama.cpp109 tok/s83.818:37Occamy-1.0-APEX-I-NanoPlus-GGUF×2llama.cpp109 tok/s82.018:01Occamy-1.0-APEX-I-NanoPlus-GGUF×2llama.cpp109 tok/s68.817:10Ornith-1.5-9B-MTP-GGUF:Q5_K_M×5llama.cpp79.1 tok/s77.2Show all 89 runs ▾Show fewer ▴
Oct 5, 2026
21:32Ornith-1.5-9B-MTP-GGUF:Q8_0×2llama.cpp58.0 tok/s63.520:24Ornith-1.5-35B-A3B-APEX-I-NanoPlus-GGUFllama.cpp93.5 tok/s81.420:13Ornith-1.5-35B-A3B-APEX-I-NanoPlus-GGUllama.cpp75.6 tok/s82.219:54Ornith-1.5-35B-A3B-APEX-I-NanoPlus-GGUF×2llama.cpp110 tok/s83.217:24Ornith-1.5-35B-A3B-GSQ-RCO-GGUF:3bit×3llama.cpp76.4 tok/s84.516:19Ornith-1.5-35B-A3B-GSQ-RCO-GGUF:3bitllama.cpp104 tok/s79.616:05Swift-1.5-Qwen3.8-27B-GSQ-RCO-GGUF:IQ3_S-mtp×3llama.cpp44.2 tok/s85.215:17Qwen3.8-27B-GSQ-RCO-GGUF:IQ3_XXS-mtp×3llama.cpp44.1 tok/s85.614:09Qwen3.8-27B-GSQ-RCO-GGUF:IQ3_XXS-mtpllama.cpp46.3 tok/s79.313:52Tiel-Coder-35B-A3B-GGUF-MTP:UD-IQ3_XXSllama.cpp88.3 tok/s84.613:34Ternary-Bonsai-2-27B-gguf:PQ2_0llama.cpp41.7 tok/s75.513:11Ternary-Bonsai-2-27B-gguf:PQ2_0×2llama.cpp43.0 tok/s74.512:46Ternary-Bonsai-2-27B-gguf:PQ2_0llama.cpp41.5 tok/s86.812:18Ternary-Bonsai-2-27B-gguf:PQ2_0×2llama.cpp41.3 tok/s74.311:57Ternary-Bonsai-2-27B-gguf:PQ2_0llama.cpp38.9 tok/s70.711:10gpt-oss-20b-GGUFllama.cpp144 tok/s67.911:07gpt-oss-20b-GGUFllama.cpp139 tok/s73.208:15Qwen3.6-35B-A3B-GGUF:UD-IQ3_XXSllama.cpp88.5 tok/s70.707:55Devstral-Small-2-24B-Instruct-2512-GGUF:UD-Q3_K_XLllama.cpp36.2 tok/s57.107:39GLM-4.7-Flash-GGUF:UD-IQ3_XXS×2llama.cpp59.8 tok/s64.507:07gpt-oss-20b-GGUFllama.cpp143 tok/s73.006:55Swift-1.5-Qwen3.8-27B-GSQ-RCO-GGUF:IQ3_S-mtpllama.cpp50.7 tok/s85.106:36Swift-1.5-Qwen3.8-27B-GSQ-RCO-GGUF:IQ3_S-mtpllama.cpp51.2 tok/s77.7Oct 4, 2026
21:18Ornith-1.5-35B-A3B-GSQ-RCO-GGUF:3bitllama.cpp88.2 tok/s84.421:08Ornith-1.5-9B-GGUF:Q8_0×2llama.cpp69.2 tok/s70.720:28gemma-4-12B-coder-fable5-composer2.5-v1-GGUF:Q4_K_Mllama.cpp56.4 tok/s56.920:12Ornith-1.5-35B-A3B-GSQ-RCO-GGUF:3bitllama.cpp101 tok/s62.217:35LiquidAI_LFM2-24B-A2B-GGUF:IQ4_XSllama.cpp185 tok/s54.217:24Tiel-Coder-35B-A3B-GGUF-MTP:UD-IQ3_XXSllama.cpp86.6 tok/s76.917:02gemma-4-26B-A4B-it-qat-GGUFllama.cpp40.4 tok/s69.116:31Qwen3-Coder-30B-A3B-Instruct:Q3_K_Sllama.cpp60.1 tok/s61.616:19Qwen3-Coder-30B-A3B-Instruct-GGUF:Q3_K_SOllama108 tok/s53.515:30Qwen3.8-27B-GSQ-RCO-GGUF:IQ3_XXS-mtpOllama34.5 tok/s81.713:58Ternary-Bonsai-2-27B-gguf-PQ2_0_Vulkanllama.cpp43.9 tok/s84.813:41Ternary-Bonsai-27B-gguf-Q2_g64llama.cpp45.4 tok/s74.312:54Ternary-Bonsai-2-27B-gguf-PQ2_0llama.cpp47.8 tok/s82.911:58gpt-oss-20b:latestOllama132 tok/s78.8