Benchmark result
IFM/K2-Horizon-0.9B-GGUF on NVIDIA GeForce RTX 4070 — 300.5 tok/s
Measured with Unsloth Studio on October 15, 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
3rd fastest of 7 runs of this model on NVIDIA GeForce RTX 4070 · median 270.2 tok/s.
sort
- 1323.4tok/s29.8Unsloth Studio · Q4_K_M · KV q8_0
- 2323.3tok/s37.4Unsloth Studio · Q4_K_M · KV q8_0
- 3300.5tok/s34.0Unsloth Studio · Q4_K_M · KV q8_0this run
- 4270.2tok/s27.1Unsloth Studio · Q4_K_M · KV q8_0
- 5261.0tok/s25.1Unsloth Studio · Q4_K_M · KV q4_0
- 6259.9tok/s20.8Unsloth Studio · Q4_K_M · KV q8_0
- 7252.4tok/s31.3Unsloth Studio · Q4_K_M · KV q8_0
Reproduce this run
toolUnsloth Studio
modelIFM/K2-Horizon-0.9B-GGUF (Q4_K_M)
context65,536
kv cacheq8_0
thinkingon
efforthigh
think limitnone
samplingtemperature 0.7 · top_p 0.95 · top_k -1 · min_p 0.01 · presence_penalty 0
clientv0.4.95+102
Set those in Unsloth Studio, then:
llm-benchmark benchmark --model "IFM/K2-Horizon-0.9B-GGUF" --tool "Unsloth 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.