Benchmark result
Tiel-Coder-35B-A3B-MTP-UD-IQ3_XXS on 2× NVIDIA GeForce RTX 5060 — 112.9 tok/s
Measured with llama.cpp b10884-434ddbbc0 on September 11, 2026.
What the model built
Open full screen →The coding scenario asks for a playable game in a single HTML file. This is exactly what the model returned, unedited; the frame only adds a line that tells this page how tall it is.
How this run compares
4th fastest of 21 runs of this model on 2× NVIDIA GeForce RTX 5060 · Multi-GPU · median 77.5 tok/s.
sort
- 1116.0tok/s86.6llama.cpp
- 2114.9tok/s86.3llama.cpp
- 3113.4tok/s83.0llama.cpp
- 4112.9tok/s80.0llama.cppthis run
- 5112.3tok/s82.2llama.cpp
- 6111.5tok/s81.7llama.cpp
- 789.7tok/s86.4llama.cpp
- 878.8tok/s84.7llama.cpp
- 978.7tok/s83.7llama.cpp
- 1078.3tok/s84.6llama.cpp
- 1177.5tok/s80.6llama.cpp
- 1277.5tok/s87.4llama.cpp
- 1377.3tok/s88.2llama.cpp
- 1477.0tok/s84.9llama.cpp
- 1577.0tok/s87.8llama.cpp
- 1676.7tok/s81.2llama.cpp
- 1776.7tok/s84.4llama.cpp
- 1876.3tok/s86.4llama.cpp
- 1975.6tok/s84.2llama.cpp
- 2074.2tok/s81.9llama.cpp
- 2143.7tok/s83.1llama.cpp
Reproduce this run
toolllama.cpp b10884-434ddbbc0
modelTiel-Coder-35B-A3B-MTP-UD-IQ3_XXS
context131,072
thinkingon
samplingtemperature 0.20000000298023224 · top_p 0.8999999761581421 · top_k 0 · min_p 0.05000000074505806 · presence_penalty 0 · repeat_penalty 1
clientv0.4.59+97
Set those in llama.cpp, then:
llm-benchmark benchmark --model "Tiel-Coder-35B-A3B-MTP-UD-IQ3_XXS" --tool "llama.cpp"The client prompts for context length and thinking mode, and for the KV cache dtype on Unsloth Studio. Sampling is left at the model default — the values above are what the tool reported using, not overrides the client sent.