Recipe

kimi-linear-48b-a3b-instruct-iq4-xs-rtx-5090-32gb-llama-cpp-tp1

kimi-linear-48b-a3b-instruct-iq4-xs-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
Engine version
10085 (b4aa7dd47)
Accelerators
1
Tensor parallel
1
Context tokens
32,768
Max concurrency
1
chat
unknown
reasoning
unknown
tools
unknown
vision
unknown

Hugging Face model card

Identity

https://huggingface.co/moonshotai/Kimi-Linear-48B-A3B-Instruct
Repository
moonshotai/Kimi-Linear-48B-A3B-Instruct
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/cmsr2sedj00z6mr01roppmbu8

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. /path/to/moonshotai_Kimi-Linear-48B-A3B-Instruct-IQ4_XS.gguf
  4. --alias
  5. kimi-linear-48b-a3b-iq4xs
  6. --host
  7. 127.0.0.1
  8. --port
  9. 8094
  10. -ngl
  11. 99
  12. -c
  13. 32768
  14. -n
  15. 8192
  16. --reasoning
  17. off
  18. --jinja
FlagValue
--model/path/to/moonshotai_Kimi-Linear-48B-A3B-Instruct-IQ4_XS.gguf
--aliaskimi-linear-48b-a3b-iq4xs
--host127.0.0.1
--port8094
-ngl99
-c32768
-n8192
--reasoningoff

Measured speed

ConcurrencyContextPrefillDecodeTTFT msStatusSweep
132,7683,733.3280.8observedkimi-linear-48b-a3b-instruct-iq4-xs-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
kimi-linear-48b-a3b-instruct-iq4-xs-rtx-5090-32gb-llama-cpp-tp1
model instance id
moonshotai-kimi-linear-48b-a3b-instruct--iq4-xs
recipe source
localmaxxing
schema version
local-ai-registry/v1
speed sweep ids
kimi-linear-48b-a3b-instruct-iq4-xs-rtx-5090-32gb-llama-cpp-tp1-sweep
status
candidate

capabilities

engine

name
llama.cpp
version
10085 (b4aa7dd47)

serving

max concurrency
1
max context tokens
32,768
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
Mean prompts 1-4 from the five-prompt RTX 5090 suite at temperature 0; prompt 5 is a four-turn session completed in 2.128 s. Prompt 2 ended at the 8192-token output limit and remained length-truncated at 16384. Peak GPU utilization 100%, peak temperature 63 C. Separate context probes at configured 131072 reached 98399 actual input tokens without OOM; near-limit decode was 226.00 tok/s.
localmaxxing · observed command
llama-server --model /path/to/moonshotai_Kimi-Linear-48B-A3B-Instruct-IQ4_XS.gguf --alias kimi-linear-48b-a3b-iq4xs --host 127.0.0.1 --port 8094 -ngl 99 -c 32768 -n 8192 --reasoning off --jinja
localmaxxing · run id
cmsr2sedj00z6mr01roppmbu8
localmaxxing · tokenized · arguments
llama-server, --model, /path/to/moonshotai_Kimi-Linear-48B-A3B-Instruct-IQ4_XS.gguf, --alias, kimi-linear-48b-a3b-iq4xs, --host, 127.0.0.1, --port, 8094, -ngl, 99, -c, 32768, -n, 8192, --reasoning, off, --jinja
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/cmsr2sedj00z6mr01roppmbu8