Qwen-AgentWorld-35B-A3B-B70-Turbo

Qwen family · 35.0B params

1recorded run

209.0best tok/s

1verified

Evidence: 1 sourced run · 1 independent source · first recorded Jul 4, 2026. What these labels mean.

What the Qwen-AgentWorld-35B-A3B-B70-Turbo numbers mean

The fastest verified result on our board is 209.0 output tok/s on an Intel Arc Pro B70 at Q5_K_M GGUF with a 30720-token context. That is about 4.8 ms per generated token — 500 tokens takes roughly 2.4 seconds. Prompt processing on the same record runs at 1925.4 tok/s, so a 2,000-token prompt is read in about 1.0 seconds.

How much VRAM does Qwen-AgentWorld-35B-A3B-B70-Turbo need?

Qwen-AgentWorld-35B-A3B-B70-Turbo has 35.0B parameters. These estimates use the same formula as the VRAM calculator: weights, plus about 15% for a moderate-context KV cache and runtime overhead, plus a small engine floor.

Quantisation	Estimated VRAM	Smallest card that fits
Q4_K_M (4-bit)	23.4 GB	32 GB
Q8_0 (8-bit)	43.6 GB	64 GB
FP16 (16-bit)	81.3 GB	96 GB

Long contexts and large batches raise the real number — read the smallest card as "possible with a reduced context". Measured throughput is on the benchmark board.

Fastest GPUs for this model (best tok/s)

Intel Arc Pro B70209.0

Models with a similar footprint

Model	Params	Q4_K_M weights
Apodex-1.1-mini	35B	19.7 GB
badtheorylabs_BTL-4	35B	19.7 GB
BTL-4-Compact	35B	19.7 GB
KAT-Coder-V2.5-Dev-AWQ-W4A16-ASYM	35B	19.7 GB
Kwaipilot_KAT-Coder-V2.5-Dev	35B	19.7 GB
MiniCPM-V 4.7-35B-A3B	35B	19.7 GB

Ranked by distance from Qwen-AgentWorld-35B-A3B-B70-Turbo's parameter count. Full benchmark board.

All benchmarks Methodology

GPU	Model	Quant	Backend	Ctx	tok/s	prompt tok/s	VRAM GB	Source
Intel Arc Pro B70 · 32 GB	Qwen-AgentWorld-35B-A3B-B70-Turbo	Q5_K_M GGUF	llama.cpp SYCL moe-ready build used by the Age…	30720	209.0 ✓	1925.4	—	src