Recipe

minimax-m2-7-autoround-w4a16-intel-arc-pro-b70-32gb-vllm-tp4

minimax-m2-7-autoround-w4a16-intel-arc-pro-b70-32gb-vllm-tp4

vllm recipe for MiniMax-M2.7-int4-AutoRound on Intel Arc Pro B70 with tensor parallelism 4.

Record

Status
candidate
Source
localmaxxing
Engine
vllm
Engine version
0.20.1-local
Accelerators
4
Tensor parallel
4
Context tokens
2,048
Max concurrency
1
chat
unknown
reasoning
unknown
tools
unknown
vision
unknown

Hugging Face model card

Identity

https://huggingface.co/Lasimeri/MiniMax-M2.7-int4-AutoRound
Repository
Lasimeri/MiniMax-M2.7-int4-AutoRound
Status
known
Link type
Exact Hub repository

Public Hugging Face repository confirmed by the Hub API.

Candidate: useful compatibility or speed evidence. The registry does not offer Run until promotion requirements are met.

Observed configuration

vllm

Evidence only · candidate · reference

Candidate evidence — not a Run contract

Source
https://www.localmaxxing.com/en/runs/cmp52gqmb000vo3015qj4kl9t

No tokenized launch fields. This is measured or documented compatibility, not a Docker launch.

Measured speed

ConcurrencyContextPrefillDecodeTTFT msStatusSweep
2,04861.8historicalminimax-m2-7-autoround-w4a16-intel-arc-pro-b70-32gb-vllm-tp4-sweep

Remaining fields

Identity, launch, related records, and measured speed are shown above. This is the rest of the normalized record.

hardware count
4
hardware id
intel-arc-pro-b70-32gb
id
minimax-m2-7-autoround-w4a16-intel-arc-pro-b70-32gb-vllm-tp4
model instance id
lasimeri-minimax-m2-7-int4-autoround--autoround-w4a16
recipe source
localmaxxing
schema version
local-ai-registry/v1
speed sweep ids
minimax-m2-7-autoround-w4a16-intel-arc-pro-b70-32gb-vllm-tp4-sweep
status
candidate

capabilities

engine

name
vllm
version
0.20.1-local

serving

max concurrency
1
max context tokens
2,048
tensor parallel
4
Provenance & metadata (3)

facts

capabilities.chat · provenance · captured at
2026-08-27T06:04:13.773Z
capabilities.chat · reason capability-not-verified
capabilities.chat · state unknown
capabilities.reasoning · provenance · captured at
2026-08-27T06:04:13.773Z
capabilities.reasoning · reason capability-not-verified
capabilities.reasoning · state unknown
capabilities.tools · provenance · captured at
2026-08-27T06:04:13.773Z
capabilities.tools · reason capability-not-verified
capabilities.tools · state unknown
capabilities.vision · provenance · captured at
2026-08-27T06:04:13.773Z
capabilities.vision · reason capability-not-verified
capabilities.vision · state unknown
description · provenance · captured at
2026-08-31T22:39:04Z
description · reason derived-from-recipe-identifiers
description · state known
engine.graph mode · provenance · captured at
2026-08-27T06:04:13.773Z
engine.graph mode · reason runtime-detail-not-published
engine.graph mode · state unknown
metadata.localmaxxing.engine flags.kvCacheDtype · provenance · captured at
2026-08-27T06:04:13.773Z
metadata.localmaxxing.engine flags.kvCacheDtype · reason not-observed
metadata.localmaxxing.engine flags.kvCacheDtype · state unknown
serving.kv cache tokens · provenance · captured at
2026-08-27T06:04:13.773Z
serving.kv cache tokens · reason kv-cache-capacity-not-published
serving.kv cache tokens · state unknown
serving.max concurrency · provenance · captured at
2026-09-01T01:34:02Z
serving.max concurrency · reason server-capacity-derived-from-source-evidence
serving.max concurrency · state known
serving.max context tokens · provenance · captured at
2026-09-01T01:30:07Z
serving.max context tokens · reason exact-live-run-context-length
serving.max context tokens · state known

metadata

localmaxxing · base lineage
known
localmaxxing · engine flags · attentionBackend
triton
localmaxxing · engine flags · flashAttn
No
localmaxxing · engine flags · gpuLayers
0
localmaxxing · engine flags · mtpEnabled
No
localmaxxing · engine flags · specDecoding
No
localmaxxing · engine flags · tensorParallel
4
localmaxxing · hardware label
4x Intel Arc Pro B70 32GB
localmaxxing · notes
MiniMax M2.7 AutoRound W4A16 on 4x Intel Arc Pro B70 via vLLM/XPU TP4. Mean of two repeat p512/n1536 runs: 61.761 and 61.744 output tok/s, 82.349 and 82.326 total tok/s. Quality smoke passed separately under the same graph/Triton recipe; no speculative decoding, no expert dropping, no power-limit increase. Uses llm-scaler INT4 MoE decode path, MiniMax attention delayed allreduce, TRITON_ATTN, and FULL_DECODE_ONLY XPU graph with non-Inductor CompilationMode.NONE because Inductor compiled decode produced invalid all-NUL output.
localmaxxing · revision unpinned
Yes
localmaxxing · run id
cmp52gqmb000vo3015qj4kl9t

provenance

captured at
2026-08-27T06:04:13.773Z

sources

captured atkindurl
2026-08-27T06:04:13.773Znormalized-recipewww.localmaxxing.com/en/runs/cmp52gqmb000vo3015qj4kl9t