Benchmark search

Try these searches

Found 12 run configurations

Each row is a run configuration (model + hardware + framework + quantization); repeated runs of the same setup are merged — raw counts appear in each result’s source line.

ModelQwen-AgentWorld-35B-A3B

Qwen-AgentWorld-35B-A3B · Q5_K_M

L0 Self-reported

NVIDIA RTX 5070 Ti · 16GB · vLLM

Other

2026-10-06

29.3 tok/s

Decode

219.8 tok/s

Prefill

0.45 s

TTFT

16 GB

VRAM

Config (latest run)
Linux CUDA 12.6 Q5_K_M vLLM 0.7.3 Flash Attention

Sources: Other · 1 runs · 1 independent sources

Qwen-AgentWorld-35B-A3B · Q4_K_M

L0 Self-reported

NVIDIA RTX 5070 Ti · 16GB · vLLM

Other

2026-10-06

35.8 tok/s

Decode

268.5 tok/s

Prefill

0.37 s

TTFT

16 GB

VRAM

Config (latest run)
Linux CUDA 12.6 Q4_K_M vLLM 0.7.3 Flash Attention

Sources: Other · 1 runs · 1 independent sources

Qwen-AgentWorld-35B-A3B · Q5_K_M

L0 Self-reported

NVIDIA RTX 5070 Ti · 16GB · llama.cpp

Other

2026-10-06

26.9 tok/s

Decode

201.8 tok/s

Prefill

0.5 s

TTFT

16 GB

VRAM

Config (latest run)
Linux CUDA 12.6 Q5_K_M llama.cpp b4200 Flash Attention

Sources: Other · 1 runs · 1 independent sources

Qwen-AgentWorld-35B-A3B · Q4_K_M

L0 Self-reported

NVIDIA RTX 5070 Ti · 16GB · llama.cpp

Other

2026-10-06

32.8 tok/s

Decode

246 tok/s

Prefill

0.41 s

TTFT

16 GB

VRAM

Config (latest run)
Linux CUDA 12.6 Q4_K_M llama.cpp b4200 Flash Attention

Sources: Other · 1 runs · 1 independent sources

Qwen-AgentWorld-35B-A3B · Q5_K_M

L0 Self-reported

NVIDIA RTX 5090 · 32GB · vLLM

Other

2026-10-06

58.6 tok/s

Decode

439.5 tok/s

Prefill

0.23 s

TTFT

26.5 GB

VRAM

Config (latest run)
Linux CUDA 12.6 Q5_K_M vLLM 0.7.3 Flash Attention

Sources: Other · 1 runs · 1 independent sources

Qwen-AgentWorld-35B-A3B · Q4_K_M

L0 Self-reported

NVIDIA RTX 5090 · 32GB · vLLM

Other

2026-10-06

71.5 tok/s

Decode

536.3 tok/s

Prefill

0.19 s

TTFT

21.8 GB

VRAM

Config (latest run)
Linux CUDA 12.6 Q4_K_M vLLM 0.7.3 Flash Attention

Sources: Other · 1 runs · 1 independent sources

Qwen-AgentWorld-35B-A3B · Q5_K_M

L0 Self-reported

NVIDIA RTX 5090 · 32GB · llama.cpp

Other

2026-10-06

53.8 tok/s

Decode

403.5 tok/s

Prefill

0.25 s

TTFT

26.5 GB

VRAM

Config (latest run)
Linux CUDA 12.6 Q5_K_M llama.cpp b4200 Flash Attention

Sources: Other · 1 runs · 1 independent sources

Qwen-AgentWorld-35B-A3B · Q4_K_M

L0 Self-reported

NVIDIA RTX 5090 · 32GB · llama.cpp

Other

2026-10-06

65.6 tok/s

Decode

492 tok/s

Prefill

0.2 s

TTFT

21.8 GB

VRAM

Config (latest run)
Linux CUDA 12.6 Q4_K_M llama.cpp b4200 Flash Attention

Sources: Other · 1 runs · 1 independent sources

Qwen-AgentWorld-35B-A3B · Q5_K_M

L0 Self-reported

NVIDIA RTX 4090 · 24GB · vLLM

Other

2026-10-06

33 tok/s

Decode

247.5 tok/s

Prefill

0.4 s

TTFT

24 GB

VRAM

Config (latest run)
Linux CUDA 12.6 Q5_K_M vLLM 0.7.3 Flash Attention

Sources: Other · 1 runs · 1 independent sources

Qwen-AgentWorld-35B-A3B · Q4_K_M

L0 Self-reported

NVIDIA RTX 4090 · 24GB · vLLM

Other

2026-10-06

40.2 tok/s

Decode

301.5 tok/s

Prefill

0.33 s

TTFT

21.8 GB

VRAM

Config (latest run)
Linux CUDA 12.6 Q4_K_M vLLM 0.7.3 Flash Attention

Sources: Other · 1 runs · 1 independent sources

Qwen-AgentWorld-35B-A3B · Q5_K_M

L0 Self-reported

NVIDIA RTX 4090 · 24GB · llama.cpp

Other

2026-10-06

30.2 tok/s

Decode

226.5 tok/s

Prefill

0.44 s

TTFT

24 GB

VRAM

Config (latest run)
Linux CUDA 12.6 Q5_K_M llama.cpp b4200 Flash Attention

Sources: Other · 1 runs · 1 independent sources

Qwen-AgentWorld-35B-A3B · Q4_K_M

L0 Self-reported

NVIDIA RTX 4090 · 24GB · llama.cpp

Other

2026-10-06

36.9 tok/s

Decode

276.8 tok/s

Prefill

0.36 s

TTFT

21.8 GB

VRAM

Config (latest run)
Linux CUDA 12.6 Q4_K_M llama.cpp b4200 Flash Attention

Sources: Other · 1 runs · 1 independent sources

12 results

Related models
Related hardware
Related frameworks
Related quantizations