model releases

All currently popular local models in one table + Opus 4.8 results

model-releasesr-localllama

If you are thinking what model will fit best your HW specs and tasks you are doing here is one table with all currently popular models that still can be considered as local. LLM Test Scores Feature De

Content type

All Blog X Post Paper YouTube Reddit GitHub

Clear filters

Model Releases

All You Need Is Non-Commutative Words

DGX agent

arXiv:2608.29314v1 Announce Type: new Abstract: We represent lexical tokens as unitary matrices and encode each sentence as their ordered product. The noncommutativity of matrix product captures word

model-releasesarxiv-cs-cl

1 Sept 2026

ALTSTEER: Selective Safety Steering for Moving Beyond Hard Refusals to Constructive Alternatives

arXiv:2608.30197v1 Announce Type: new Abstract: Safety alignment is essential for deploying large language models, requiring systems to prevent harmful compliance while preserving helpfulness on benig

Amortized Anchor Refinement for Deployable Continuous-Time 4D Gaussian Reconstruction

arXiv:2608.30218v1 Announce Type: new Abstract: Continuous-time 4D reconstruction remains impractical on standalone XR headsets. Per-scene optimization demands deployment-infeasible compute, and lower

An Agentic Retrobiosynthesis Framework with Learned Frontier Selection

arXiv:2608.30702v1 Announce Type: cross Abstract: Large language models are increasingly used as agents for multistep retrosynthesis, raising the question of how much their search policy contributes i

An Explainable Coherence Score for Detecting Temporal Inconsistencies in Political News

arXiv:2608.29175v1 Announce Type: new Abstract: Temporal inconsistencies, such as mandates attributed outside their real interval, events presented as past before they occurred, or inverted causal seq

An Open-Source, Event-Driven Pipeline for Cryptocurrency Market Data: Ingestion, Forecasting, and On-Chain Fraud Detection

arXiv:2608.29973v1 Announce Type: new Abstract: Cryptocurrency markets generate high-frequency, multi-source data that is expensive to work with unless a team already has commercial-grade streaming an

Analytic Dynamics: Learning Physics-Grounded Representation for Fast Intrinsic Dynamics Inference from Monocular Videos

arXiv:2608.31025v1 Announce Type: new Abstract: Inferring object dynamics from visual observations is essential for intelligent agents to reason about and interact with the physical world, yet remains

AnnotateAnything: Automatic Annotation of 3D Assets for Robot Manipulation

arXiv:2606.17446v2 Announce Type: replace-cross Abstract: Simulation enables scalable robot data collection, but raw 3D assets provide only geometry, lacking the semantic, interactive, and physical kn

Annotated Surrogate Retrieval for Polish Statutory Law

arXiv:2608.30929v1 Announce Type: new Abstract: We present a family of retrieval methods for Polish statutory law built on document surrogates: language-model annotations attached to statutory article

Another Update: in openai, we ARE continuously working on Prism, the scientific/technical writing surface. It is owned by a small team; impr…

Another Update: in openai, we ARE continuously working on Prism, the scientific/technical writing surface. It is owned by a small team; improvements are coming, although we wish we can ship more. The

model-releasesopenai--x

Anthropic launches Claude Fable 5.1 after inking $35B cloud deal with Lambda

model-releasessiliconangle

Anthropic PBC today debuted Claude Fable 5.1 and Claude Mythos 5.1, its most capable large language models to date. The launch comes a day after the company inked a 35 billion infrastructure deal with

Anthropic launches Claude Fable 5.1 and says it’s up to 45 percent cheaper for agentic work

model-releasesthe-verge-ai

Anthropic says its newest AI models, Fable 5.1 and Mythos 5.1, address criticisms from customers about price, data retention, and overzealous safeguards. The company claims Claude Fable 5.1 offers str

Anthropic says Fable 5.1 sets new standards on coding, knowledge work, and long-running problem-solving tasks, and can fix the root causes of software issues (Carl Franzen/VentureBeat)

Carl Franzen / VentureBeat: Anthropic says Fable 5.1 sets new standards on coding, knowledge work, and long-running problem-solving tasks, and can fix the root causes of software issues — It's only th

model-releasestechmeme

Anthropic says Fable 5.1 will cost an estimated 25% less than Fable 5 for typical workloads and up to 45% less for highly agentic work (Zac Hall/9to5Mac)

Zac Hall / 9to5Mac: Anthropic says Fable 5.1 will cost an estimated 25% less than Fable 5 for typical workloads and up to 45% less for highly agentic work — In June, Anthropic first released Fable 5,

model-releasestechmeme

Anthropic unveils Claude Fable 5.1, which is generally available, and Mythos 5.1, for trusted partners and with cybersecurity and life sciences safeguards (Anthropic)

Anthropic: Anthropic unveils Claude Fable 5.1, which is generally available, and Mythos 5.1, for trusted partners and with cybersecurity and life sciences safeguards — We're introducing Claude Fable 5

model-releasestechmeme

AnyWorld: Factorized Egocentric World Models for Cross-Embodiment Generalization

model-releasesarxiv-cs-ro

arXiv:2608.29242v1 Announce Type: new Abstract: Collecting contact-rich robot experiences at scale remains a major bottleneck for generalizable manipulation. Beyond data quantity, robot learning also

APIFlow-Bench: Measuring Whether Agents Survive Long, Dependent API Workflows

arXiv:2608.29128v1 Announce Type: new Abstract: Tool-using agents are commonly evaluated by a single bit: whether an end-to-end workflow completed. This metric fails to distinguish failures that matte

APT: Atomic Physical Transitions for Causal Video-Language Understanding

arXiv:2606.18586v2 Announce Type: replace-cross Abstract: Physical events are not understood by their names alone, but by the causal state changes that compose them. A clip-level label such as 'bounce

Are LLMs Ready to Assist Physicians? PhysAssistBench for Interactive Doctor-Patient-EHR Assistance

arXiv:2606.18613v4 Announce Type: replace-cross Abstract: The most plausible near-term role of medical LLMs is to assist rather than replace physicians, yet current evaluations often test isolated cap

Are Online Skill and Memory Modules Always Worth Their Tokens? A Budget-Constrained Study of Web Agents

arXiv:2606.15017v2 Announce Type: replace Abstract: Online web agents often augment a base actor with memory, workflow, or skill modules. These modules can improve performance, but they also consume t

Argument-Aware Semantic Alignment of Normative Texts: A Toulmin-Based Neuro-Symbolic Approach

arXiv:2608.29529v1 Announce Type: cross Abstract: Semantic alignment between specialized normative texts is challenging when equivalent requirements use different terms, syntax, and levels of abstract

Arkios: An Open Bilingual English-Nepali Language Model Trained From Scratch, with a Devanagari-Aware Tokenizer

arXiv:2608.30092v1 Announce Type: cross Abstract: We present Arkios, a 1.04B-parameter dense transformer pretrained from scratch on 150B tokens of bilingual English-Nepali text, using a custom single-

Aspire: Can Models Self-Evolve from Vague Goals?

arXiv:2608.31111v1 Announce Type: new Abstract: Many important forms of human learning begin with a vague goal, such as 'become a better physicist' or 'improve at research.' Learners must interpret th

Auditing MCQA Benchmarks through Probability Landscapes

arXiv:2608.30372v1 Announce Type: new Abstract: As Large Language Models rapidly advance, performance on standard multiple-choice question answering (MCQA) benchmarks is reaching saturation. While the

Automated Researchers Can Reliably Mitigate Alignment Failures

arXiv:2608.28945v1 Announce Type: new Abstract: Automating alignment research may accelerate progress toward aligned AI, but whether it does is hard to measure. Luckily, many alignment failures, such

Autoregressive Mosaics: Probing 2D Spatial Reasoning in Text-Only Language Models

arXiv:2608.30751v1 Announce Type: new Abstract: Large language models (LLMs) trained only on text and code can sometimes generate programs that draw recognizable images. However, it is unclear whether

A.X K2 Technical Report

arXiv:2608.30181v1 Announce Type: new Abstract: We introduce A.X K2, a 688B-parameter Mixture-of-Experts (MoE) language model trained from scratch as a high-performance foundation for agentic applicat

b10730

qwen4exp: sum the indexer heads by slices (#28023) qwen4exp: sum the indexer heads by slices The head reduction went through a transpose and a sum_rows over ne[1], which left sum_rows with ne0 = 4, on

b10731

qwen4exp: support recurrent state rollback (#28123) MTP speculative decoding needs the target state to move back by the number of rejected draft tokens. Without rollback support the context is classif

b10733

ci: Bump ggml-org/ccache-action to v1.2.24 (#28083) Website: https://llama.app Attestations: https://github.com/ggml-org/llama.cpp/attestations/44375174 macOS/iOS: macOS Apple Silicon (arm64) macOS Ap

b10734

metal: enable Metal 4.0 tensor API on M5+/A19+ (#27461) metal : request Metal 4.0 language version for the tensor API metal : load the tensor API kernels from a separate metallib tests : add external-

b10736

tests : fix log verbosity for test-llama-archs (#28147) tests : fix log verbosity for test-llama-archs cont : naming cont : add note Website: https://llama.app Attestations: https://github.com/ggml-or

b10737

qwen4exp: fix seq_cp, block position keying, mtmd input, cuda abort, add tests (#27941) qwen4exp: follow up fixes -kvu NaN collapse fix Assisted-by: Claude indexer cache ext.x/ext.y restore fix Assist

b10738

sycl : support limit max alloc memory within 2GB for host-pinned memory (#27559) Website: https://llama.app Attestations: https://github.com/ggml-org/llama.cpp/attestations/44438607 macOS/iOS: macOS A

b10739

metal : add fa-vec tuning for M2 Max (#28015) Rows for M2 Max (30 GPU cores) collected with 'ggml-metal-tuning fa-vec --dtype f16,q8_0', pasted into fa_vec_tuned_table. ref: https://github.com/ggml-or

b10740

metal : fix more leaks due to missing autoreleasepools (#27883) metal : fix more leaks due to missing autoreleasepools metal : rename variable metal : fix another missing pool warning Co-authored-by:

b10741

model : load hparams.n_layer_nextn before n_layer() calls (#28159) load hparams.n_layer_nextn before n_layer() calls remove duplicate loads Website: https://llama.app Attestations: https://github.com/

b10742

metal : add fa-vec tunings for A18 Pro (MacBook Neo) (#28152) Website: https://llama.app Attestations: https://github.com/ggml-org/llama.cpp/attestations/44474754 macOS/iOS: macOS Apple Silicon (arm64

b10743

metal : add fa-vec tuning for M2 Pro (#28122) metal: add fa-vec tuning for M2 Pro metal : update fa-vec tuning for M2 Pro with new dtypes Website: https://llama.app Attestations: https://github.com/gg

b10749

context : autoscale n_ctx_train when yarn scaling specified (#28030) Website: https://llama.app Attestations: https://github.com/ggml-org/llama.cpp/attestations/44499127 macOS/iOS: macOS Apple Silicon

b10750

kv-cells: look up the n-gram history in the sequence position index (#28040) get_prev_tokens() rebuilt a (seq, pos) -> token hash map on every ubatch by walking all used cells, while llama_kv_cells al

b10751

cuda: fuse MoE weighted expert reduction (#25952) cuda : fuse MoE weighted reduction (mul + view + add) The MoE combine tail currently writes weighted expert outputs to global memory before reducing t

Balance of Benchmarks: Semantic Density Reweighting for Benchmark Multiplicity and Task-Conditioned Evaluation

arXiv:2608.30044v1 Announce Type: new Abstract: Language models are commonly compared by averaging scores across a benchmark list with equal weight. Such lists grow through publication outside an expl

Balancing Privacy, Utility, and Safety in LLM Alignment through Preference Optimization

model-releasesarxiv-cs-lg

arXiv:2608.30141v1 Announce Type: cross Abstract: Preference optimization is widely used to align large language models with human preferences, but preference-data composition may also influence priva

Banger paper from the Qwen team. If you evaluate agents on anything longer than a single session, this one is worth your time. (bookmark it)…

Banger paper from the Qwen team. If you evaluate agents on anything longer than a single session, this one is worth your time. (bookmark it) E-Commerce Bench runs an agent through a simulated 365-day

model-releasesdair-ai--x

Bayesian-Optimized Superpixel-GrabCut for Traceable Optic Disc Segmentation

arXiv:2608.29196v1 Announce Type: new Abstract: Optic disc (OD) segmentation is essential for diagnosing ophthalmic pathologies from retinal fundus images. However, prevailing deep learning approaches

Behavior-Skill: A Fine-Grained Benchmark for Evaluating Vision-Language-Action Policies in Long-Horizon Tasks

model-releasesarxiv-cs-ro

arXiv:2608.30536v1 Announce Type: new Abstract: Reliable execution of long-horizon mobile manipulation tasks remains challenging because overall task success depends on the successful completion of mu