Recipe

gpt-oss-20b-mxfp4-apple-m5-max-128gb-mlx-tp1

gpt-oss-20b-mxfp4-apple-m5-max-128gb-mlx-tp1

JSON API Find compatible recipes

Observed LocalMaxxing leaderboard run. Evidence for compatibility, not an executable launch contract.

Record

Status: candidate
Source: localmaxxing
Engine: mlx
Engine version: mlx-vlm 0.6.3
Accelerators: 1
Tensor parallel: 1
Context tokens: 2,048
Max concurrency: 16
chat: unknown
reasoning: unknown
tools: unknown
vision: unknown

Hugging Face model card

Identity

https://huggingface.co/openai/gpt-oss-20b

Repository: openai/gpt-oss-20b
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

mlx

Evidence only · candidate · reference

Candidate evidence — not a Run contract

Source: https://www.localmaxxing.com/en/runs/cmqszh407012sqr01oer5egep

Observed source tokens

Mechanical split of the source command. Unverified against the engine CLI. The registry does not offer Run for this recipe.

mlx_vlm.server
--model
mlx-community/gpt-oss-20b-MXFP4-Q8
--port
8125
--max-kv-size
16384
#
16
concurrent
streams,
256
max_tokens
each,
temperature
0.7,
continuous
batching
(default),
no
prompt
cache

Flag	Value
`--model`	`mlx-community/gpt-oss-20b-MXFP4-Q8`
`--port`	`8125`
`--max-kv-size`	`16384`

Measured speed

Concurrency	Context	Prefill	Decode	TTFT ms	Status	Sweep
16	2,048	—	388.7	—	observed	gpt-oss-20b-mxfp4-apple-m5-max-128gb-mlx-tp1-sweep

Remaining fields

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

hardware count: 1
hardware id: apple-m5-max-128gb
id: gpt-oss-20b-mxfp4-apple-m5-max-128gb-mlx-tp1
model instance id: openai-gpt-oss-20b--mxfp4
recipe source: localmaxxing
schema version: local-ai-registry/v1
speed sweep ids: gpt-oss-20b-mxfp4-apple-m5-max-128gb-mlx-tp1-sweep
status: candidate

capabilities

engine

name: mlx
version: mlx-vlm 0.6.3

serving

max concurrency: 16
max context tokens: 2,048
tensor parallel: 1

Provenance & metadata (3)

facts

capabilities.chat · provenance · captured at: 2026-08-30T09:10:02Z

capabilities.chat · reason capability-not-verified

capabilities.chat · state unknown

capabilities.reasoning · provenance · captured at: 2026-08-30T09:10:02Z

capabilities.reasoning · reason capability-not-verified

capabilities.reasoning · state unknown

capabilities.tools · provenance · captured at: 2026-08-30T09:10:02Z

capabilities.tools · reason capability-not-verified

capabilities.tools · state unknown

capabilities.vision · provenance · captured at: 2026-08-30T09:10:02Z

capabilities.vision · reason capability-not-verified

capabilities.vision · state unknown

engine.graph mode · provenance · captured at: 2026-08-30T09:10:02Z

engine.graph mode · reason runtime-detail-not-published

engine.graph mode · state unknown

serving.kv cache tokens · provenance · captured at: 2026-08-30T09:10:02Z

serving.kv cache tokens · reason kv-cache-capacity-not-published

serving.kv cache tokens · state unknown

metadata

localmaxxing · hardware label: M5 Max
localmaxxing · notes: Aggregate end-to-end throughput across 16 concurrent streams (256 max_tokens each, distinct prompts so no prefix-cache reuse), mlx-vlm 0.6.3 server continuous batching. Weights: mlx-community/gpt-oss-20b-MXFP4-Q8 (native MXFP4). tokSOut = aggregate output tok/s; per-stream median ~24 tok/s. Reproducible (382-389 across two runs). Peak unified memory ~12.3 GB (RSS). Apple M5 Max / 128GB, macOS 27.0. Single-stream (bs=1) ~126 tok/s.
localmaxxing · observed command: mlx_vlm.server --model mlx-community/gpt-oss-20b-MXFP4-Q8 --port 8125 --max-kv-size 16384 # 16 concurrent streams, 256 max_tokens each, temperature 0.7, continuous batching (default), no prompt cache
localmaxxing · run id: cmqszh407012sqr01oer5egep
localmaxxing · tokenized · arguments: mlx_vlm.server, --model, mlx-community/gpt-oss-20b-MXFP4-Q8, --port, 8125, --max-kv-size, 16384, #, 16, concurrent, streams,, 256, max_tokens, each,, temperature, 0.7,, continuous, batching, (default),, no, prompt, cache
localmaxxing · tokenized · fidelity: faithful

provenance

captured at: 2026-08-30T09:10:02Z

sources

captured at	kind	url
2026-08-30T09:10:02Z	normalized-recipe	www.localmaxxing.com/en/runs/cmqszh407012sqr01oer5egep ↗