Unsloth Studio benchmarks
An inference engine for running open-weight LLMs locally.
8 community reports
Unsloth Studio results by GPU
Hardware with reported Unsloth Studio results, with each report's model, quant and speed, newest first. Runs on several cards, with speculative decoding, with batched requests or with part of the model in system RAM say so, since each describes a different setup. Prompt processing is how fast the input is read, not how long you wait for the first word. How to measure time to first token.
Unsloth Studio on NVIDIA RTX 5070 Ti Laptop 12GB2 reports
- Qwen3.5 9B · NPU27.9 t/s
- Qwen3.8 27B · UD-Q4_K_XL4.5 t/s(speculative, part in system RAM)
Unsloth Studio on M1 16GB1 report
- Gemma 4 8B · 4bit8.2 t/s
Unsloth Studio on NVIDIA RTX 3060 12GB1 report
- DeepSeek V4 Flash 284B · 13B active · IQ2_M3.5 t/s(2 cards, part in system RAM)
Unsloth Studio, GPU not identified1 report
- Qwen3.8 27B · IQ4_XS33.0 t/s
This engine doesn't yet have an editorial profile on llamaperf. The community reports below show how it's been used in practice across different hardware.
Compare GPUs running Unsloth Studio
Every card is compared on one model, DeepSeek V4.1 Flash 552B · 16B active at 3-bit, the size and bit class that most cards here have plain runs of. A card with no run of that model shows none rather than a figure from a different model.
| GPU | VRAM | Reports | Median speed, DeepSeek V4.1 Flash 552B · 16B active 3-bit |
|---|---|---|---|
| NVIDIA RTX 5070 Ti Laptop 12GBnvidia | 12GB | 2 | no DeepSeek V4.1 Flash 552B · 16B active 3-bit run reported |
| AMD RX 9070amd | 16GB | 1 | no DeepSeek V4.1 Flash 552B · 16B active 3-bit run reported |
| M1 16GBapple | 16GB | 1 | no DeepSeek V4.1 Flash 552B · 16B active 3-bit run reported |
| M3 Ultra 256GBapple | 256GB | 1 | Gen 40.0 t/s |
| NVIDIA RTX 3060 12GBnvidia | 12GB | 1 | no DeepSeek V4.1 Flash 552B · 16B active 3-bit run reported |
| NVIDIA RTX 5060 Ti 16GBnvidia | 16GB | 1 | no DeepSeek V4.1 Flash 552B · 16B active 3-bit run reported |
| NVIDIA RTX 5090nvidia | 32GB | 1 | no DeepSeek V4.1 Flash 552B · 16B active 3-bit run reported |
Unsloth Studio against other engines
Pairs of plain runs on the same card, of the same model size at the same bit class: one device, one request, no speculative decoding, the whole model in memory. Context length and build still differ between the two sides, and each side shows its own.
No matched pair yet. No card has plain runs of one model size at one bit class on Unsloth Studio and on another engine, so llamaperf can't say how it compares on speed. Add a run.
Top models on Unsloth Studio
Frequently asked
Is Unsloth Studio faster than other engines?
llamaperf has no matched comparison for Unsloth Studio yet: no card has plain runs of the same model size at the same bit class on Unsloth Studio and on another engine. Speed claims about engines need that pairing, so this page doesn't make one.