Skip to content
Benchmarks / LLM /Entry / FP8

LLM · Entry · FP8

A 4B model at FP8 on 12 GB. FP8 needs Ada or newer. For an Ampere card such as an RTX 3090, use the INT4 lane.

Best Decode
tok/s
Ranked results
0
0 distinct systems
GPU models
0
Minimum VRAM
12 GB
~14 min run

Best result per GPU

Decode · tok/s
No results yet for this view.
Show the values in this chart (0 rows)
GPUGPUsSamplesBest (tok/s)Mean (tok/s)

Download as CSV· Full open dataset

No results match these filters.

What is pinned

The profile fixes all values below. LocalMax checks them at each submission. If the runtime flags do not agree, it publishes the result but does not rank it.

Model
Repository
Qwen/Qwen3-4B-Instruct
Revision
pending freeze
Precision
fp8_e4m3
Parameters
4 B
Licence
Apache-2.0
Runtime
Engine
vllm 0.26.0
Harness
aiperf
dtype
auto
max-model-len
16384
gpu-memory-utilization
0.9
max-num-seqs
16
enforce-eager
false
disable-log-requests
true
tensor-parallel-size
1
swap-space
0
Workloads
interactive
input_tokens=1024 output_tokens=256 concurrency=1 input_source=fixed_synthetic input_seed=20260801
concurrency
input_tokens=1024 output_tokens=256 concurrency_sweep=1/2/4/8 input_source=fixed_synthetic input_seed=20260801
longcontext
input_tokens=8192 output_tokens=1024 concurrency=1 input_source=fixed_synthetic input_seed=20260801
Ranking
Ranked on
Decode (tok/s, higher is better)
Gate
TTFT p95 ≤ 3000
Gate
Inter-token p95 ≤ 200
Gate
Error rate ≤ 0
Also shown
Prefill, TTFT p50, TTFT p95, Inter-token p95, Peak throughput @c8, Energy / token, Model load
Design notes
  • This profile is a release candidate. The runtime is pinned. The model revision is not final. This leaderboard is provisional.
  • The value for gpu-memory-utilization is 0.9, not 0.95. A 12 GB card often drives a display.
  • Decode and prefill are separate values. Decode follows memory bandwidth. Prefill follows compute.
  • The Entry tier accepts single-GPU results only. On a multi-GPU system the runner uses one device.