Recipe

deepseek-v4-flash-0731-ad-iq1-m-radeon-ai-pro-r9700-32gb-llama-cpp-tp3

deepseek-v4-flash-0731-ad-iq1-m-radeon-ai-pro-r9700-32gb-llama-cpp-tp3

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

Record

Status
candidate
Source
localmaxxing
Engine
llama.cpp
Engine version
b10275
Accelerators
3
Tensor parallel
3
Context tokens
32,768
Max concurrency
1
chat
unknown
reasoning
unknown
tools
unknown
vision
unknown

Hugging Face model card

Identity

https://huggingface.co/deepseek-ai/DeepSeek-V4-Flash-0731
Repository
deepseek-ai/DeepSeek-V4-Flash-0731
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/cmsgbo9jl014fpp01hvmzqiif

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. DeepSeek-V4-Flash-0731-AD-IQ1_M.gguf
  4. -md
  5. dspark-DeepSeek-V4-Flash-0731-MXFP4.gguf
  6. --spec-type
  7. draft-dspark
  8. --spec-draft-n-max
  9. 5
  10. -ngl
  11. 99
  12. --flash-attn
  13. auto
  14. --no-mmap
  15. -ts
  16. 0.8/1.05/1.15
  17. -ctk
  18. f16
  19. -ctv
  20. f16
  21. -c
  22. 32768
  23. -np
  24. 1
FlagValue
-mDeepSeek-V4-Flash-0731-AD-IQ1_M.gguf
-mddspark-DeepSeek-V4-Flash-0731-MXFP4.gguf
--spec-typedraft-dspark
--spec-draft-n-max5
-ngl99
--flash-attnauto
-ts0.8/1.05/1.15
-ctkf16
-ctvf16
-c32768
-np1

Measured speed

ConcurrencyContextPrefillDecodeTTFT msStatusSweep
132,76826248.51,870.1observeddeepseek-v4-flash-0731-ad-iq1-m-radeon-ai-pro-r9700-32gb-llama-cpp-tp3-sweep

Remaining fields

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

hardware count
3
hardware id
radeon-ai-pro-r9700-32gb
id
deepseek-v4-flash-0731-ad-iq1-m-radeon-ai-pro-r9700-32gb-llama-cpp-tp3
model instance id
deepseek-ai-deepseek-v4-flash-0731--ad-iq1-m
recipe source
localmaxxing
schema version
local-ai-registry/v1
speed sweep ids
deepseek-v4-flash-0731-ad-iq1-m-radeon-ai-pro-r9700-32gb-llama-cpp-tp3-sweep
status
candidate

capabilities

engine

name
llama.cpp
version
b10275

serving

max concurrency
1
max context tokens
32,768
tensor parallel
3
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
rocm
localmaxxing · hardware label
Radeon AI Pro R9700
localmaxxing · notes
Run 4/10 of a full 10-run distribution (submitting all runs, no cherry-pick). Code-generation prompt, temp 0.6, 256 tok out, fresh prefill (no prompt cache). DSpark speculative decode, external MXFP4 draft, draft n-max 5. Eco power profile (-75mV, 250W/GPU cap). Median of the 10: 42.4 tok/s decode.
localmaxxing · observed command
llama-server -m DeepSeek-V4-Flash-0731-AD-IQ1_M.gguf -md dspark-DeepSeek-V4-Flash-0731-MXFP4.gguf --spec-type draft-dspark --spec-draft-n-max 5 -ngl 99 --flash-attn auto --no-mmap -ts 0.8/1.05/1.15 -ctk f16 -ctv f16 -c 32768 -np 1
localmaxxing · run id
cmsgbo9jl014fpp01hvmzqiif
localmaxxing · tokenized · arguments
llama-server, -m, DeepSeek-V4-Flash-0731-AD-IQ1_M.gguf, -md, dspark-DeepSeek-V4-Flash-0731-MXFP4.gguf, --spec-type, draft-dspark, --spec-draft-n-max, 5, -ngl, 99, --flash-attn, auto, --no-mmap, -ts, 0.8/1.05/1.15, -ctk, f16, -ctv, f16, -c, 32768, -np, 1
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/cmsgbo9jl014fpp01hvmzqiif