Recipe

deepseek-v4-flash-q2-apple-m5-max-128gb-llama-cpp-tp1

deepseek-v4-flash-q2-apple-m5-max-128gb-llama-cpp-tp1

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

Record

Status
candidate
Source
localmaxxing
Engine
llama.cpp
Engine version
b10079
Accelerators
1
Tensor parallel
1
Context tokens
4,096
Max concurrency
1
chat
unknown
reasoning
unknown
tools
unknown
vision
unknown

Hugging Face model card

Identity

https://huggingface.co/huihui-ai/Huihui-DeepSeek-V4-Flash-abliterated-ds4-GGUF
Repository
huihui-ai/Huihui-DeepSeek-V4-Flash-abliterated-ds4-GGUF
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

llama.cpp

Evidence only · candidate · reference

Candidate evidence — not a Run contract

Source
https://www.localmaxxing.com/en/runs/cms0nfewo009rs2011l53fkr5

Observed source tokens

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

  1. llama-server
  2. -m
  3. Huihui-DeepSeek-V4-Flash-BF16-abliterated-ds4-Q2.gguf
  4. --host
  5. 127.0.0.1
  6. --port
  7. 8089
  8. -ngl
  9. 99
  10. -c
  11. 4096
  12. --flash-attn
  13. on
  14. --temp
  15. 0
FlagValue
-mHuihui-DeepSeek-V4-Flash-BF16-abliterated-ds4-Q2.gguf
--host127.0.0.1
--port8089
-ngl99
-c4096
--flash-attnon
--temp0

Measured speed

ConcurrencyContextPrefillDecodeTTFT msStatusSweep
14,0963,544.812.1150.9observeddeepseek-v4-flash-q2-apple-m5-max-128gb-llama-cpp-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
deepseek-v4-flash-q2-apple-m5-max-128gb-llama-cpp-tp1
model instance id
huihui-ai-huihui-deepseek-v4-flash-abliterated-ds4-gguf--q2
recipe source
localmaxxing
schema version
local-ai-registry/v1
speed sweep ids
deepseek-v4-flash-q2-apple-m5-max-128gb-llama-cpp-tp1-sweep
status
candidate

capabilities

engine

name
llama.cpp
version
b10079

serving

max concurrency
1
max context tokens
4,096
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 · backend
metal
localmaxxing · hardware label
M5 Max
localmaxxing · notes
DeepSeek-V4-Flash abliterated (Q2, 81GB on disk) on Apple M5 Max 128GB unified memory. llama.cpp b10079 with Metal backend, all layers on GPU, flash attention. 535-token prompt, 128 output tokens, temperature=0, 2 warmup + 5 measured runs (median). Consistent: 11.95-12.37 tok/s. Some DeepSeek-specific fused ops (Lightning Indexer, HC pre/comb/post) fall back to CPU on Metal; native Metal support would improve throughput. Demonstrates a 180B-class MoE fitting in 128GB unified memory.
localmaxxing · observed command
llama-server -m Huihui-DeepSeek-V4-Flash-BF16-abliterated-ds4-Q2.gguf --host 127.0.0.1 --port 8089 -ngl 99 -c 4096 --flash-attn on --temp 0
localmaxxing · run id
cms0nfewo009rs2011l53fkr5
localmaxxing · tokenized · arguments
llama-server, -m, Huihui-DeepSeek-V4-Flash-BF16-abliterated-ds4-Q2.gguf, --host, 127.0.0.1, --port, 8089, -ngl, 99, -c, 4096, --flash-attn, on, --temp, 0
localmaxxing · tokenized · fidelity
faithful

provenance

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

sources

captured atkindurl
2026-08-30T09:10:02Znormalized-recipewww.localmaxxing.com/en/runs/cms0nfewo009rs2011l53fkr5