Menu
Back previous/next report: k j

Intel 4x Arc(TM) Pro B70

Submitted by Intelon 2026-06-26

Result statement

Intel ran GPT-OSS 120B on 4x Arc(TM) Pro B70 and achieved a verified 1,546.8 tokens/second System TPS at 26,998.5ms TTFT P90 in the Standardized division of the Available category .

[1] MLPerf Endpoints v1.0 GPT-OSS 120B; retrieved from endpoints.mlcommons.org on October 22, 2026. Result —, run 7: concurrency = 384 queries, TTFT P90 = 26,998.5ms, System TPS = 1,546.8 tokens/sec and interactivity = 4.0 tokens/sec per user. Result verified by MLCommons Association. The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information.

Submission Info
Intel
c14186d2-0438-4b2b-a204-6c66dfbd5365
Submission Status PUBLISHED
2026-06-26
—
Division Standardized
Deployment Client on Premises
Model
Model Name GPT-OSS 120B
Division Standardized
Model Precision FP16
Model Link —
Calibration Link —
Model Notes —
Datasets
Dataset Name GPT-OSS Dataset
Dataset Type performance
Average Input Tokens 5011
Dataset Link —

System TPS vs. Tokens/s per User

5001.0k1.5kSystem TPS5101520Tokens/s per User

System TPS vs. Concurrency

5001.0k1.5kSystem TPS0100200300400# of queries

Time to First Token P90 vs. Concurrency

010k20kTTFT P90 (ms)0100200300400# of queries

Tokens/s per User vs. Concurrency

5101520Tokens/s per User0100200300400# of queries
SUT Summary
Intel
4x Arc(TM) Pro B70
Available
System Category Datacenter · —
4x Intel(R) Arc(TM) Pro B70 Graphics
1
1
Max Supported Concurrency 1,536
Node Type ID 0
Node Count 1
Software
Ubuntu
Framework vLLM 0.1.dev11435+gfa77e1a63
vLLM 0.1.dev11435+gfa77e1a63
Not detected: CUDA/ROCm/XPU runtime not found
Not available
Link to Container —
Operating System linux debian ubuntu 24.04
Filesystem ext4 vfat
Other Software —
Software Notes —
Hardware

Processor

Intel(R) Xeon(R) 698X
1
86
VCPUs Per Processor 172

Accelerator

Intel(R) Arc(TM) Pro B70 Graphics
4
GDDR6
32GiB
Not available
PCIe 5.0 x16

Host / Storage

134G
Memory Configuration 8 x 16GB DDR5-6400
Storage Capacity 1.8 TB NVMe SSD
Storage Type NVMe SSD

Networking

3x Ethernet
Ethernet
Network Topology —

Other

Other Hardware —
Cooling —
Hardware Notes —
Run Data
Field NameRun 1Run 2Run 3Run 4Run 5Run 6Run 7
Run Date06/26/202606/26/202606/26/202606/26/202606/26/202606/26/202606/26/2026
Concurrency16324064128256384
System Tokens/Second311.5545.4663.1937.81,326.31,522.51,546.8
Tokens/Second per User19.517.016.614.710.45.94.0
TTFT P90 (ms)1041.81110.61137.11339.81766.63104.226998.5
Utilization20.1%35.3%42.9%60.6%85.7%98.4%100.0%
Configuration Summarygpt-oss-120b-benchmark · online · c16gpt-oss-120b-benchmark · submission · c32gpt-oss-120b-benchmark · submission · c40gpt-oss-120b-benchmark · submission · c64gpt-oss-120b-benchmark · submission · c128gpt-oss-120b-benchmark · submission · c256gpt-oss-120b-benchmark · submission · c384
Config Typeonlinesubmissionsubmissionsubmissionsubmissionsubmissionsubmission
Config Region———————
Target QPS———————
Result Statecompletecompletecompletecompletecompletecompletecomplete
Measured Accuracy Score———————
Benchmark Versionba95c2f7251d707251d707251d707251d707251d707251d70
Result Version0.1.00.1.00.1.00.1.00.1.00.1.00.1.0
Archive URIgs://mlperf-endpoints-blob/runs/0e3682e7-5b5e-4dbb-a068-e68c380b7e00/archive.tar.gzgs://mlperf-endpoints-blob/runs/4a4bce79-4897-4223-aa6f-f166131df2c6/archive.tar.gzgs://mlperf-endpoints-blob/runs/85a2b623-000c-4010-9884-b02f484112f4/archive.tar.gzgs://mlperf-endpoints-blob/runs/8fa8a7df-685f-4df5-ab05-df89fbd772f0/archive.tar.gzgs://mlperf-endpoints-blob/runs/81faee26-b00c-4e49-b44d-612dc48538dc/archive.tar.gzgs://mlperf-endpoints-blob/runs/a30b386b-77ee-4a1f-baf6-7016f0b1ee15/archive.tar.gzgs://mlperf-endpoints-blob/runs/bd5d352a-2ef9-4499-8373-412cd7724cd1/archive.tar.gz
QPS0.22910.39450.48310.69080.97801.12871.1512
Total Output Tokens8,698,2048,841,0048,779,5178,682,5558,673,4388,627,6378,594,475
Average Output Tokens1,359.941,382.271,372.661,357.51,356.071,348.911,343.73
Run Duration (s)27,922.4416,211.4913,240.599,258.686,539.635,666.915,556.15
Total Queries6,3966,3966,3966,3966,3966,3966,396
Total Queries Issued6,3966,3966,3966,3966,3966,3966,396
Total Queries Failed0000000
Time to First Token (TTFT) (ms)
Minimum216.7216.0217.4239.4290.2397.1642.1
Average678.8748.9813.5999.21748.74505.620619.4
P50629.7662.4675.5715.2949.71482.214702.6
P901041.81110.61137.11339.81766.63104.226998.5
P951210.71312.91362.51605.12172.44430.736237.6
P991603.51782.71878.22888.635788.4108954.3194077.1
P9995433.211943.319032.231341.568319.3143341.6233837.9
Maximum8185.515214.222701.135584.572180.0148949.9239378.9
Time Per Output Token (TPOT) (ms)
Minimum46.250.348.150.349.760.966.9
Average50.858.059.667.294.3162.4230.1
P5050.857.959.667.394.6164.3235.1
P9051.959.561.269.397.9168.4238.8
P9552.260.161.869.998.9169.5240.0
P9952.861.162.971.1101.0172.2242.3
P99953.662.765.072.9104.1177.2245.0
Maximum54.8114.4113.174.1108.5190.3250.0
Query Latency (ms)
Minimum30577.81383.01063.234035.647246.171126.396318.8
Average69731.180857.282551.992084.2129490.3223274.1329335.2
P5068583.078453.380475.990583.4127444.4219880.2324416.1
P9086677.9100534.8103337.2115851.3162988.7283233.2418366.6
P9594001.0108760.4110007.0124458.1175818.0306836.1451690.7
P99108874.9130471.9127699.9141476.7204148.0360063.9539161.6
P999128584.3589166.0607952.9173235.0240468.8411631.2613361.8
Maximum520748.0593328.8617914.4192981.5323771.5477606.0709762.9