model releases

ggml-cuda: HIP: optimize packed byte subtraction (__vsubss4 -> __vsub4) (#29478) ggml-cuda: HIP: optimize non-saturating packed byte subtraction (__vsubss4) CI: ignore 1 spilled vgpr in fattn_vec Co-a

hexagon: optimize concat op (#29673) hex-concat: reduce pkts in gather/transpose hot loop gather directly into dst buffer, use special instruction for gather sync hex-concat: use fastdiv replace calls

model : support classifier_pooling for rerankers (#29627) model : support classifier_pooling for ModernBERT rerankers Assisted-by: Claude Opus 5.5 model : read classifier pooling type in load_hparams

ggml : check row bounds in get_rows_back (#29575) Signed-off-by: Adrien Gallouët angt@huggingface.co Website: https://llama.app Attestations: https://github.com/ggml-org/llama.cpp/attestations/5139241

gguf : reject tensor size that wraps after padding (#26979) GGML_PAD(nbytes, alignment) wraps to 0 when nbytes is within (alignment - 1) of SIZE_MAX, which silently bypassed the size overflow guard in

Hexagon f16 activation ops (#29209) hexagon: add F16 support for activation ops (SILU/GELU/GELU_QUICK/GEGLU/SWIGLU) Widens ggml_hexagon_supported_activations() to accept F16 (src0/dst/src1 must agree

ggml-zdnn: impl buffer reset, fix memory leaks (#29637) cont: fix code style Signed-off-by: Aaron Teo aaron.teo1@ibm.com Website: https://llama.app Attestations: https://github.com/ggml-org/llama.cpp/

vendor: update BoringSSL to 0.20260929.0 (#29669) Website: https://llama.app Attestations: https://github.com/ggml-org/llama.cpp/attestations/51432581 macOS/iOS: macOS Apple Silicon (arm64) macOS Appl

add GLM-5.3-Flash (GLM5-Next) support (#27773) Rebase GLM-Next support onto master, and migrate to llama-memory-hybrid-idx Add initial MTP support Merge branch optimizations. Reduce allocated compute

SYCL: reduce tensor allreduce sync with pinned host buffers (#29604) Website: https://llama.app Attestations: https://github.com/ggml-org/llama.cpp/attestations/51446793 macOS/iOS: macOS Apple Silicon

musa : define CUDA_ARCH for device passes (#29508) The MUSA vendor header never defined CUDA_ARCH, so every architecture test in the shared ggml-cuda sources evaluated to 0. Kernel bodies gated on the

openvino: serve GET_ROWS on a weight view from the base Constant (#28381) openvino: serve GET_ROWS on a weight view from the base Constant Resolve view_src when collecting weight Constants so a view o

cpu: accept BF16 in src1 of mul_mat (#28937) cpu: accept BF16 in src1 of mul_mat ggml_conv_1d_dw builds its im2col in F32 when the kernel is BF16, then calls ggml_mul_mat(im2col, kernel), which puts F

ggml : add BF16 unary, GLU, binary and scale ops (CPU, CUDA) (#29675) ggml : add BF16 unary, GLU, binary and scale ops (CPU, CUDA) ggml-cpu : use per-op _bf16 functions for BF16 unary and GLU ops Assi

llama: llama_prefetch_rows (#29599) llama: llama_prefetch_rows llama: support row prefetch on Windows Apply the Windows port contributed by @praneshgo unchanged. Source: #29599 (comment) avoid exposin

ci : fix Models Backend Check by shortening the hrm_text fixture (#29744) The fixture recycles its two blocks over 8 cache slots, so the fp16 error builds up past the 1e-4 NMSE bound on the Vulkan T4

jinja : support coerced array attributes (#29574) support coerced array attributes add tests Website: https://llama.app Attestations: https://github.com/ggml-org/llama.cpp/attestations/51590476 macOS/

mimo : support dflash (convert + feature extraction) (#29650) convert : update to support dflash cont : fix Co-authored-by: Sigbjørn Skjæret sigbjorn.skjaeret@huggingface.co Co-authored-by: Sigbjørn S

cli: exit on stdin EOF and drop the console wide Ctrl+C broadcast (#29722) cli: exit on stdin EOF and drop the console wide Ctrl+C broadcast On Windows the simple input reader sends CTRL_C_EVENT to ev

ggml/gguf : fix integer overflow (#29384) ggml: fix integer overflow guard for zero-element tensors ggml: validate number of elements in tensor to prevent integer overflow ggml: fix error print Websit

glm5-next: give dead indexer slots unique scatter rows (#29745) The sparse indexer mask is built with a set_rows scatter. Padded pools, absent sequences and missing tail cells all pointed to the same

batch: migrate the rest of examples to llama_batch_ext (#29601) migrate the rest test-thread-safety rm common_batch_staged Website: https://llama.app Attestations: https://github.com/ggml-org/llama.cp

llama: properly handle KV on training (#28520) llama: properly handle KV on training improve Website: https://llama.app Attestations: https://github.com/ggml-org/llama.cpp/attestations/51644581 macOS/

test-llama-archs : toggle causal_attn to catch graph shape changes (#29724) After the device decode, flip causal_attn off, decode n_ubatch/2 then n_ubatch tokens. Both have the same node count, so a s

'AREX-2 is a 27B-parameter long-horizon agent model from the Beijing Academy of Artificial Intelligence (BAAI). It learns to improve a solution over multiple test-time rounds: propose, measure, reflec

arXiv:2609.36738v1 Announce Type: cross Abstract: Optimizer momentum is usually stored as a parameter-sized moving average of past gradients, which makes history costly and fixes each past signal in t

arXiv:2609.37041v1 Announce Type: cross Abstract: Self-Distillation Fine-Tuning (SDFT) enables a language model to act as its own teacher: by conditioning on a demonstration, the model produces an imp

arXiv:2609.37400v1 Announce Type: new Abstract: Generating realistic 3D dance from music is a challenging task that requires accurate synchronization with musical rhythms while capturing the spatial c

arXiv:2609.36920v1 Announce Type: new Abstract: Comprehensive evaluations of automatic speech recognition (ASR) for Iberian languages remain limited, and low-resource languages, biases, and efficiency

​ Building a local voice chat pipeline: Struggling with fast, natural PT-BR conversation models I'm currently trying to build a local AI voice chat system that can use tools via an LLM. I'm testing tw

arXiv:2609.36097v1 Announce Type: cross Abstract: Using predictive models to explain cognition requires more than accurate behavioral predictions. Input ablations offer an appealing route: remove info

arXiv:2609.36628v1 Announce Type: new Abstract: Vision-Language Models (VLMs) can generate rich video captions, yet often misidentify which person performs an action or which limb is involved, particu

arXiv:2609.37066v1 Announce Type: cross Abstract: Post-training is central to mathematical reasoning in modern large language models (LLMs), but endpoint pass@1 alone underidentifies what has changed.

arXiv:2609.37905v1 Announce Type: cross Abstract: Click-Through Rate prediction, a core task in recommendation and advertising systems, relies on modeling interactions among sparse categorical feature

arXiv:2609.36598v1 Announce Type: new Abstract: A video can exhibit convincing motion and photorealism yet fail immediately when visual text collapses. Unlike generic scene content, visual text is unf

arXiv:2609.38019v1 Announce Type: new Abstract: We present RGOR (Reference-Grounded Oral Refinement), an audio-driven lip-sync framework that renders the mouth of the specific person being dubbed rath

arXiv:2609.37027v1 Announce Type: new Abstract: Low-Rank Adaptation (LoRA) is a widely used approach to parameter-efficient fine-tuning (PEFT), yet a performance gap can remain relative to full fine-t

arXiv:2602.11678v2 Announce Type: replace Abstract: Multimodal Large Language Models (MLLMs) have shown remarkable progress in visual understanding, yet they suffer from a critical limitation: structu

arXiv:2609.37377v1 Announce Type: new Abstract: On-policy distillation (OPD) trains students using teacher feedback on their own sampled responses, yet how prompt choice shapes transfer across teacher

arXiv:2609.35841v1 Announce Type: cross Abstract: Mutation testing evaluates test-suite adequacy by injecting synthetic faults into program code. However, traditional rule-based tools often generate l

arXiv:2609.36888v1 Announce Type: new Abstract: Mixed human-LLM documents require locating authorship transitions from detector scores whose reliability varies across text units. Existing weighted mea

arXiv:2609.35875v1 Announce Type: new Abstract: Multi-agent debate (MAD) reportedly improves reasoning and factuality over single-model inference, but prior work treats agents as symmetric peers, leav

arXiv:2609.35831v1 Announce Type: new Abstract: Modern large language models now support context windows of more than one million tokens, which has raised the question of whether retrieval-augmented g

arXiv:2609.37713v1 Announce Type: new Abstract: Existing product matching benchmarks primarily contain English-language product data and are often dominated by a single product category, such as elect

arXiv:2609.37993v1 Announce Type: cross Abstract: The BITEM team entered both subtasks of the NTCIR-19 R2C2 task with a single agentic pipeline, in which a model searches, reads and records evidence o

arXiv:2512.07011v2 Announce Type: replace-cross Abstract: Modern large language models increasingly require long contexts for reasoning and multi-document tasks, but attention's quadratic complexity c