llama.cpp
add GLM-5.3-Flash (GLM5-Next) support
#27773
Merged
Go
Login via GitHub
Home
Pricing
FAQ
Install
Login
via GitHub
Overview
Commits
61
Changes
View On
GitHub
add GLM-5.3-Flash (GLM5-Next) support
#27773
am17an
merged 61 commits into
ggml-org:master
from
timkhronos:GLM5.3-Flash
github-actions
added
model
github-actions
added
mtmd
github-actions
added
conversion
ngxson
commented on 2026-08-27
ngxson
commented on 2026-08-27
ngxson
commented on 2026-08-27
Rebase GLM-Next support onto master, and migrate to llama-memory-hybr…
9370c82d
timkhronos
force pushed
to
9370c82d
54 days ago
Add initial MTP support
d4ce8259
Merge branch optimizations. Reduce allocated compute buffer size, spe…
d2fc716c
Review driven changes, remove env vars, protect tensors
98557115
Strip MTP for initial PR
b93ed51f
Clean up after mtp strip
ba436fd3
Clean up after mtp strip
3ff8d92c
Update speculative.cpp
f176c8a3
Update llama-context.h
c34db581
Clean up after mtp strip
0aa327b3
Fix tokenizer ignore merges
ec1cdbf0
Improve quantization protection selection
7152e9bf
timkhronos
marked this pull request as ready for review
52 days ago
timkhronos
requested a review
52 days ago
timkhronos
requested a review
from
CISC
52 days ago
timkhronos
requested a review
from
ggerganov
52 days ago
ngxson
commented on 2026-08-30
Refactor mhc helpers, graph base
fdc54bed
Merge branch 'master' into GLM5.3-Flash
85439412
ngxson
approved these changes on 2026-08-31
Lint Fixes
74bb0e39
CISC
commented on 2026-08-31
Apply suggestions from code review
a771613a
Skip glm5-next in model saver, fix CRLF
5728a4bf
Skip glm5-next in sweep
c35bddd0
timkhronos
requested a review
from
JohannesGaessler
50 days ago
github-actions
added
testing
Remove T4 fallback
3bdb2d84
CISC
commented on 2026-09-01
Review cleanup
81f95af4
CISC
commented on 2026-09-01
Merge branch 'master' into GLM5.3-Flash
3498cc75
Merge branch 'ggml-org:master' into GLM5.3-Flash
611e4070
Review suggestions
1eca274e
CISC
approved these changes on 2026-09-01
Defer separate MTP gguf handling to MTP PR, drop filter
7de5a8e3
ggerganov
commented on 2026-09-01
Repad n_head_kv
9fe9fd71
kpool init apply
a386cd76
Merge branch 'ggml-org:master' into GLM5.3-Flash
5b05fccb
Order by descending score
120eb9e4
CISC
commented on 2026-09-02
Drop guard
d0c55891
ggerganov
commented on 2026-09-02
read kpool from hparams, clarify kpool cache flags, remove kpool_buil…
8c289399
Add glm5-next support to model saver and add arch test fixture
a2f1d20c
ggerganov
commented on 2026-09-02
ggerganov
commented on 2026-09-02
ggerganov
commented on 2026-09-02
ggerganov
commented on 2026-09-02
Review cleanup
fe3187de
Kpool pooled caching clarify
2b533e09
Add multi stream support
ff6be954
Merge branch 'ggml-org:master' into GLM5.3-Flash
ad74ec35
Finish Rebase
1b564d2d
Sparse FA fir DSA prefill
99fdaab4
Merge branch 'ggml-org:master' into GLM5.3-Flash
db54719f
Const
5c4bd50b
Merge branch 'master' into GLM5.3-Flash
cc05bfee
Update llama-model.cpp to fix rebase error
8134115f
CISC
requested a review
from
ggerganov
44 days ago
fairydreaming
commented on 2026-09-09
Merge remote-tracking branch 'upstream/master' into GLM5.3-Flash
05f7c98a
gguf-py : merge tensor map entries for HC tensors
ae9893ff
model : use build_gdn_l2_norm in GLM5_NEXT implementation
3f238d44
chore : remove trailing whitespace
26ace38c
model : use new OP precision setting API in GLM5_NEXT implementation
de253435
am17an
commented on 2026-09-12
am17an
commented on 2026-09-12
mtmd : use ggml_swiglu_clamp in GLM5V and apply the image token limit
df40ac60
llama : keep the GLM5-Next k-pool layout across ubatches
166c62d5
model : write the GLM5-Next recurrent rollback checkpoints
a22c8459
Merge master into GLM5.3-Flash (PR #27773): resolve merge conflicts
36f338f1
llama: fix PR #27773 test-save-load-state restore failure
93aefbcf
llama: fix PR #27773 gpu-rocm graph reallocation
c2ed4857
llama : fix GLM5-Next k-pool layout staleness after edits and shared …
1bceb45d
am17an
approved these changes on 2026-09-28
ggerganov
assigned
ggerganov
22 days ago
llama : fix build_attn_mha stream stride for non-contiguous q
18217f60
llama : re-derive GLM5-Next k-pool sharing on state_read/state_drop
8583e617
quant : drop duplicate GLM5-Next hc_ filter
2ad2809d
Merge branch 'master' into GLM5.3-Flash (PR #27773)
544765ce
ggerganov
commented on 2026-09-29
glm5-next: scope K-pool cache access to indexed operations
4cab1fbf
ggerganov
commented on 2026-09-29
glm5-next: keep K-pool access in hybrid index memory
3746f394
glm5-next: keep mHC graph builders model-local
8aaabc0b
glm5-next: mark only touched pools per ubatch
1665c0e6
ggerganov
approved these changes on 2026-09-30
am17an
merged
649dcb10
into master
20 days ago
ggerganov
added
highlight
Login to write a write a comment.
Login via GitHub
Reviewers
ggerganov
am17an
CISC
ngxson
pwilkin
fairydreaming
JohannesGaessler
Assignees
ggerganov
Labels
model
testing
mtmd
conversion
highlight
Milestone
No milestone
Login to write a write a comment.
Login via GitHub