Benchmark result
swift-1.5-qwen3.8-27b-uncensored on NVIDIA GeForce RTX 5060 Ti +1 more — 279.0 tok/s
Measured with ninfer manual on October 7, 2026.
Support llm-bench.io
llm-bench.io is free. If it helped you, I'd appreciate your support to cover the infrastructure costs.
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
12th fastest of 19 runs of this model on NVIDIA GeForce RTX 5060 Ti +1 more · Multi-GPU · median 282.2 tok/s.
sort
- 1302.2tok/s79.8ninfer
- 2298.6tok/s85.2ninfer
- 3296.0tok/s79.1ninfer
- 4295.1tok/s84.2ninfer
- 5294.2tok/s84.5ninfer
- 6293.0tok/s84.2ninfer
- 7291.0tok/s85.3ninfer
- 8284.5tok/s78.4ninfer
- 9284.1tok/s83.3ninfer
- 10282.2tok/s85.1ninfer
- 11279.7tok/s82.3ninfer
- 12279.0tok/s87.9ninferthis run
- 13277.7tok/s83.3ninfer
- 14276.6tok/s86.8ninfer
- 15276.6tok/s83.0ninfer
- 16276.2tok/s80.7ninfer
- 17274.1tok/s84.5ninfer
- 18270.2tok/s86.6ninfer
- 19267.3tok/s83.7ninfer
Reproduce this run
toolninfer manual
modelswift-1.5-qwen3.8-27b-uncensored
context100,000
thinkingon
clientv0.4.89+102
Set those in ninfer, then:
llm-benchmark benchmark --model "swift-1.5-qwen3.8-27b-uncensored" --tool "ninfer"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.