model: add Kimi-K3 text model - #26185

Merged

ngxson merged 22 commits into

ggml-org:masterfrom

pwilkin:kimi-k3-text

Aug 15, 2026

Commits on Aug 11, 2026

model: add Kimi-K3 text model
pwilkin
committed
901f9b3
model: fix ty errors in the Kimi-K3 converter
pwilkin
committed
d171e72
Update conversion/kimi_k3.py

pwilkin
and
worthant
committed
f9181dc
Increase LLAMA_MAX_EXPERTS from 512 to 1024
pwilkin
committed
28a84b1
tests : support for Kimi K3 in archs test

sszymczy
authored and
pwilkin
committed
50a5431
chat : add Kimi K3 chat format (reasoning, content, typed tool calls)

authored and
pwilkin
committed
03efbfe
chat : add message_delimiters for Kimi K3

authored and
pwilkin
committed
2524866
fix: apply nits from @ngxson and text fixes from @danielhanchen
pwilkin
committed
cf8b10b
tests : added missing hyperparameters and tensors for Kimi K3 in test-llama-archs

sszymczy
authored and
pwilkin
committed
0302fa3
chore : move overly verbose header file comments to Kimi K3 source file

sszymczy
authored and
pwilkin
committed
c495421
tests : re-enabled KIMI_K3 in test-llama-archs for WebGPU backend

sszymczy
authored and
pwilkin
committed
4bb78d5
model-saver : emit kda_gate_lower_bound for Kimi K3

SolshineCode
authored and
pwilkin
committed
0d5346b

Commits on Aug 13, 2026

Refactor conditional for model architecture check
pwilkin
authored
3f36269

Commits on Aug 14, 2026

tests : re-enabled (again) KIMI_K3 and MINIMAX_M3 in test-llama-archs for WebGPU backend
sszymczy
committed
e881427

Commits on Aug 15, 2026

fix code comments
ngxson
committed
3f99a41
add template on conversion
ngxson
committed
493ad3b
move repack_mxfp4_blocks to model base
ngxson
committed
e1039c2
nits
ngxson
committed
7b90295
add_value_length
ngxson
committed
29f5f9c
optimize res_stack construction
ngxson
committed
d8980d6
Merge branch 'master' into kimi-k3-text
ngxson
committed
3080e61
nits
ngxson
committed
914db7f