Benchmark result
mistralai/ministral-3-3b on NVIDIA GeForce RTX 2050 — 8.2 tok/s
Measured with LM Studio on October 16, 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
5th fastest of 5 runs of this model on NVIDIA GeForce RTX 2050 · median 26.5 tok/s.
sort
- 126.7tok/s57.1LM Studio · Q4_K_M
- 226.6tok/s50.9LM Studio · Q4_K_M
- 326.5tok/s46.9LM Studio · Q4_K_M
- 414.7tok/s45.3LM Studio · Q4_K_M
- 58.2tok/s50.4LM Studio · Q8_0this run
mistralai/ministral-3-3b on other hardware
- 26.9tok/s58.2NVIDIA RTX 500 Ada Generation Laptop GPU · Q4_K_M
- 13.1tok/s46.9AMD Radeon Vega Series / Radeon Vega Mobile Series
- 13.0tok/s43.0AMD Radeon Vega Series / Radeon Vega Mobile Series
- 12.9tok/s47.9AMD Radeon Vega Series / Radeon Vega Mobile Series
- 12.7tok/s40.4AMD Radeon Vega Series / Radeon Vega Mobile Series
Reproduce this run
toolLM Studio
modelmistralai/ministral-3-3b (Q8_0)
context8,192
clientv0.4.96+102
Set those in LM Studio, then:
llm-benchmark benchmark --model "mistralai/ministral-3-3b" --tool "LM Studio"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.