Recipe

deepseek-v4-flash-0731-ud-iq1-s-rtx-5090-32gb-llama-cpp-tp1

deepseek-v4-flash-0731-ud-iq1-s-rtx-5090-32gb-llama-cpp-tp1

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

Record

Status
candidate
Source
localmaxxing
Engine
llama.cpp
Accelerators
1
Tensor parallel
1
Context tokens
2,048
Max concurrency
1
chat
unknown
reasoning
unknown
tools
unknown
vision
unknown

Hugging Face model card

Identity

https://huggingface.co/unsloth/DeepSeek-V4-Flash-0731-GGUF
Repository
unsloth/DeepSeek-V4-Flash-0731-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/cmt04563z0ehpms01an9c50gz

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. --model
  3. DeepSeek-V4-Flash-0731-UD-IQ1_S-00001-of-00003.gguf
  4. --ctx-size
  5. 65536
  6. --n-gpu-layers
  7. 999
  8. --n-cpu-moe
  9. 34
  10. --cache-type-k
  11. q8_0
  12. --cache-type-v
  13. q8_0
  14. --flash-attn
  15. on
  16. --jinja
  17. --threads
  18. 8
FlagValue
--modelDeepSeek-V4-Flash-0731-UD-IQ1_S-00001-of-00003.gguf
--ctx-size65536
--n-gpu-layers999
--n-cpu-moe34
--cache-type-kq8_0
--cache-type-vq8_0
--flash-attnon
--threads8

Measured speed

ConcurrencyContextPrefillDecodeTTFT msStatusSweep
12,048107.118233.4observeddeepseek-v4-flash-0731-ud-iq1-s-rtx-5090-32gb-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
rtx-5090-32gb
id
deepseek-v4-flash-0731-ud-iq1-s-rtx-5090-32gb-llama-cpp-tp1
model instance id
unsloth-deepseek-v4-flash-0731-gguf--ud-iq1-s
recipe source
localmaxxing
schema version
local-ai-registry/v1
speed sweep ids
deepseek-v4-flash-0731-ud-iq1-s-rtx-5090-32gb-llama-cpp-tp1-sweep
status
candidate

capabilities

draft launch

accelerator backend
nvidia
arguments
-hf, unsloth/DeepSeek-V4-Flash-0731-GGUF, --n-gpu-layers, 999, --host, 0.0.0.0, --port, 8080, -c, 2048
container port
8,080
environment · LLAMA CACHE
/root/.cache/huggingface
host port
8,080
image
ghcr.io/ggml-org/llama.cpp:server-cuda12-b10481@sha256:b2497f8834f5ecb4e38530f6bf2734b8e0be107f0f0857e259672d1cb85b71c2
ipc
host
kind
docker
shm size
16g
synthesized · generated at
2026-08-31T22:12:17Z
synthesized · image provenance
gemma-4-12b-q4-k-m-rtx-3060-12gb-llama-cpp-tp1
synthesized · template
llama-cpp-server-v1

mounts

read onlytarget
No/root/.cache/huggingface

engine

name
llama.cpp

serving

max concurrency
1
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 · backend
cuda
localmaxxing · hardware label
RTX 5090
localmaxxing · notes
llama-server with the 77 GB UD-IQ1_S (unsloth dynamic 1-bit) shards, --n-gpu-layers 999 but 34 of 43 MoE layers' experts held in system RAM (--n-cpu-moe 34), all dense/attention weights on GPU. Long-context configuration: 65,536-token context with the KV cache quantized to 8-bit on both keys and values (--cache-type-k/v q8_0) — the KV savings pay for doubled context rather than extra GPU expert layers. Flash attention on, 8 CPU threads, greedy decode (temp 1.0/top-p 1.0 measured at temperature 0 via the benchmark client). Decode is bound by CPU-side expert streaming through DDR5, not the GPU.
localmaxxing · observed command
llama-server --model DeepSeek-V4-Flash-0731-UD-IQ1_S-00001-of-00003.gguf --ctx-size 65536 --n-gpu-layers 999 --n-cpu-moe 34 --cache-type-k q8_0 --cache-type-v q8_0 --flash-attn on --jinja --threads 8
localmaxxing · run id
cmt04563z0ehpms01an9c50gz
localmaxxing · tokenized · arguments
llama-server, --model, DeepSeek-V4-Flash-0731-UD-IQ1_S-00001-of-00003.gguf, --ctx-size, 65536, --n-gpu-layers, 999, --n-cpu-moe, 34, --cache-type-k, q8_0, --cache-type-v, q8_0, --flash-attn, on, --jinja, --threads, 8
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/cmt04563z0ehpms01an9c50gz