wrapped·Jul 08, 2026
/r/r_o636l3cc-rryour run
glm-4.7-flash
on RTX 4090
129.9
tok/sdecode
RTX 4090 (24GB)
rank in tier
27/68RTX 4090 runstop 40%
best workload
agent-trace
where the rig flew
slowest workload
—
single-workload run
backend
ollamaollama is llama.cpp under the hood; pure llama.cpp tends to nudge a few % faster.
faster than
- ornith-1.5-iq4xs on AMD Ryzen 7 9800X3D 8-Core Processor125.3 tok/s
- Qwen3-8B-4bit on M3 Ultra120.5 tok/s
- DeepSeek-R1-0528-Qwen3-8B-MLX-4bit on M3 Ultra116.8 tok/s