Machine
RX 7900 XTX · Ryzen 7 9800X3D #8d2270c0
58 runs of 27 models since Oct 1, 2026. Last run Oct 30, 2026.
58
runs
27
models
32
configs
6
configs run 2+ times
Hardware
- GPU
- AMD Radeon RX 7900 XTX 24 GB
- CPU
- AMD Ryzen 7 9800X3D 8-Core Processor 8 cores
- RAM
- 31 GB
- OS
- Windows 11 Pro 26H2
- Tools
- LM Studio 44 · llama.cpp 12 · llama.cpp+NNL-LoRA 2
- Client
- 0.4.85
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
Cyber-Tiel-Coder-35B-A3B-MTP-UD-IQ3_XXS
1 run · 1 config
qwen3.8-27b-paoai-rocmfp4-strix-balanced
1 run · 1 config
tiel-coder-35b-a3b
1 run · 1 config
cyber-tiel-coder-35b-a3b-mtp
9 runs · 3 configs
★ bestIQ3_XXSLM Studio80k ctxthinking on
×6 runs
83.180.7–86.0
steady
120 tok/s
90.5–125
Cyber-Tiel-Coder-35B-A3B.APEX-I-MiniPlus-V2.1
1 run · 1 config
tiel-coder-35b-a3b-mtp
1 run · 1 config
swift-qwen3.8-27b
1 run · 1 config
Cyber-Tiel-Coder-35B-A3B.APEX-I-NanoPlus
6 runs · 3 configs
★ bestLM Studio64k ctxthinking on
×4 runs
80.779.5–86.1
steady
116 tok/s
111–120
qwen3.6-35b-a3b-mtp
1 run · 1 config
Qwen3.6-35B-A3B-UD-IQ2_XXS
12 runs · 1 config
★ bestLM Studio64k ctxthinking on
×12 runs
77.373.9–81.2
steady
210 tok/s
195–226
Compare with another machine
Run history
Oct 20, 2026
15:27Ternary-Bonsai-2-2B-gguf:PQ2_0llama.cpp326 tok/s35.215:21Ternary-Bonsai-2-2B-Q2_0llama.cpp328 tok/s26.915:12qwen2.5-coder-0.5b-instruct-q4_k_mllama.cpp357 tok/s5.815:08qwen2.5-0.5b-instruct-q4_k_m×2llama.cpp260 tok/s8.814:45qwen2.5-0.5b-nnl-distilled-clean-q4_k_m×3llama.cpp345 tok/s7.414:33teacher35b_quick_v1_q4_k_mllama.cpp336 tok/s5.914:32qwen2.5-0.5b-instruct-q4_k_m×2llama.cpp293 tok/s8.512:54nnl-qwen2.5-0.5b-repair-v2llama.cpp+NNL-LoRA245 tok/s9.312:51nnl-qwen2.5-0.5b-chatgpt-lora-v1llama.cpp+NNL-LoRA255 tok/s8.912:41qwen2.5-0.5b-instruct-q4_k_mllama.cpp341 tok/s10.0Oct 2, 2026
13:26Qwen3.6-35B-A3B-UD-IQ2_XXS×12LM Studio210 tok/s77.312:05Qwen3.6-35B-A3B-UD-IQ4_NLLM Studio182 tok/s74.411:26Cyber-Tiel-Coder-35B-A3B-MTP-UD-IQ3_XXSLM Studio114 tok/s85.810:48Cyber-Tiel-Coder-35B-A3B.APEX-I-NanoPlus×3LM Studio115 tok/s79.909:59Cyber-Tiel-Coder-35B-A3B.APEX-I-NanoPlusLM Studio107 tok/s78.509:47Cyber-Tiel-Coder-35B-A3B.APEX-I-NanoPlusLM Studio120 tok/s86.109:28Cyber-Tiel-Coder-35B-A3B.APEX-I-MiniPlus-V2.1LM Studio44.7 tok/s82.809:11Cyber-Tiel-Coder-35B-A3B.APEX-I-NanoPlusLM Studio110 tok/s82.408:37tiel-coder-35b-a3b-mtpLM Studio123 tok/s82.8Oct 1, 2026
22:39ornith-1.5-35b-a3b-heretic-mtp-apexLM Studio100 tok/s65.122:30tiel-coder-35b-a3bLM Studio55.1 tok/s83.422:21ayncoding-qwen2.5-coder-1.5bLM Studio263 tok/s14.821:19omnicoder-9bLM Studio88.3 tok/s67.419:40qwen3.8-27b-paoai-rocmfp4-strix-balancedLM Studio40.2 tok/s85.419:15lfm2.5-2.6b-qwen3.8-turbo-brilliance-power-x12-neo-maxLM Studio195 tok/s55.018:58cyber-tiel-coder-35b-a3b-mtpLM Studio122 tok/s84.018:22cyber-tiel-coder-35b-a3b-mtp×4LM Studio123 tok/s82.917:36qwen3.6-35b-a3b-mtpLM Studio122 tok/s79.703:58cyber-tiel-coder-35b-a3b-mtp×2LM Studio114 tok/s73.203:39cyber-tiel-coder-35b-a3b-mtpLM Studio114 tok/s82.002:49swift-qwen3.8-27bLM Studio29.8 tok/s81.502:22gemma-4-26b-a4b-it-rocmfp4LM Studio125 tok/s65.202:10cyber-tiel-coder-35b-a3b-mtpLM Studio118 tok/s86.001:52ornith-1.5-35b-a3b-rocmfp4LM Studio104 tok/s64.901:05spark-x2.5-4b-rocmfp4-strix_leanLM Studio184 tok/s58.200:44ornith-1.5-35b-a3b-rocmfp4LM Studio142 tok/s59.9