llamaperf

Qwen3.5 4B Fonzie0.2

on NVIDIA RTX 4070 Ti · llama.cpp · 65,536 ctx

Tone: mixedNVIDIA hardware
Oct 31, 2026
Throughput
118.0 t/s gen
Quant
Q4_K_M (GGUF)
KV cache
q8_0
VRAM reported
12 GB

Use cases

tool-useagenticmultilingualvision

Summary

User reports Fonzie0.2-4B, a LoRA fine-tune of Qwen3.5-4B, decoding about 118 tok/s single-user on an RTX 4070 Ti with llama-server, and about 157 tok/s with MTP 2. Setup is llama.cpp's llama-server with the Q4_K_M GGUF (2.8 GB plus a 0.7 GB vision projector) and a q8_0 KV cache; at 64k context the whole server takes about 5.4 GiB. The model was trained on one DGX Spark with Unsloth on 1,390 conversations in 97 minutes. The user reports accuracy gains in three agent apps but flat accuracy on everyday questions, more invented links (5.9% vs 0.3%), and that GRPO gains in their own harness did not transfer to real apps.

Compare this run

Medians of plain runs: one device, one request, no speculative decoding, the whole model in the device's memory. The report above is one setup, so read it beside them and not as one of them.

Qwen3.5 4B at 4-bit by hardware

Bars are medians on one scale, 0 to 118.0 tokens/s. Press ▶ to watch text arrive at a row's speed.

Models most run on the NVIDIA RTX 4070 Ti

Bars are medians on one scale, 0 to 118.0 tokens/s. Press ▶ to watch text arrive at a row's speed.

Everything reported on the NVIDIA RTX 4070 Ti →Every Qwen3.5 report →Estimate Qwen3.5 on this card →