Mistral Large 4 ships as open weights for coding and cyber
The 1.05T model has 52B active parameters and a 1M context; Mistral's ranking claims are narrower than independent scores.
64
- Mentions · 7d4
- Mentions · 30d26
- All mentions52
- First seenAug 30, 2026
- Last mentioned1d ago
Timeline
52 articles · newest firstMoE channel selection works without calibration data, paper finds
A weight-only reader keeps 98.1% of full accuracy at 17.8% of routed-expert reads and speeds MoE blocks versus vLLM.
42
vLLM constrained decoding lifts Kimi tool-call schema accuracy to 100%
Serving Kimi K2.6 on vLLM 0.28 matched the expected schema only 92% of the time, the authors report.
46
Small open-weight coding workers trail direct coding on checks
In one author's studies, a Qwen3-Coder worker passed 0/45 tasks while direct coding passed 45/45 on the same suite.
46
DeepSeek open-sources DeepInfer, its production inference engine
DeepSeek claims 155 trillion tokens a day and 96% cache hits; KV cache and load balancing are not fully released.
62
Google open-sources TorchTPU for running PyTorch on TPUs
TPU support also landed in vLLM, SGLang, TorchTitan, Ray, and Hugging Face libraries, with two pieces still in preview.
58
Muse Spark Open, the Most Capable US Open‑Weight Model to Date
Meta released Muse Spark Open, an 816B multimodal open-weight model with 42B active parameters and a 1M-token context.
91
LFM2.5-24B-A2B: Built for fast on-device agents
Liquid AI released open-weight LFM2.5-24B-A2B, a 24B MoE with 2B active parameters for on-device agents.
70
Google Releases Gemini 4 Argon API With 1M Token Output Limit
Google opened Gemini 4 Argon on the paid API, a sparse MoE model with a combined 1M-token window.
93
Gemini 4 Argon: Built to solve your hardest problems
Google released Gemini 4 Argon globally, ranking first on the Vals Index.
96
Gemini 4 Argon: Built to solve your hardest problems
Google launched frontier model Gemini 4 Argon, ranked first on the Vals Index.
90
DuoDiscover searches optimizations with cheap proxy evaluators
Authors say it doubles expert speedup on SWE-fficiency and reports 1.92 times vLLM and 1.38 times SQLite.
52
StartLux-Decision: AutoThink Decision Models with Reinforcement Learning for Calibrated Deliberative Decisions
StartLux-Decision models from 0.8B to 27B add confidence-guided reasoning trained with RLCD².
50
autotrust/JEV-35B-NVFP4 — new model trending #21 on Hugging Face
AutoTrust released JEV-35B-NVFP4, a 4-bit quantization of its 35B decision model that cuts GPU memory by 65%.
48
vllm-sr/d3 — new model trending #24 on Hugging Face
vLLM Semantic Router released d3, a 27B multimodal model that returns calibrated decisions instead of generated text.
55
Metasploit Wrap Up: A Collection of What Can Only Be Called Eclectic Modules
Rapid7's Metasploit wrap-up adds 12 modules, including unauthenticated RCE for N-central, Langflow, LiteLLM, and dizqueTV.
52
nvidia/Qwen3.8-Flash-Next-FP4-Local — new model trending #30 on Hugging Face
NVIDIA published an FP4-quantized Qwen3.8-Flash-Next checkpoint for local Blackwell inference, occupying about 96 GiB.
56
Critical LMCache RCE Vulnerability Remains Unpatched, Public PoC Exploit Available
Unpatched LMCache flaw CVE-2026-105192 allows unauthenticated remote code execution, and JFrog published a PoC.
74
Unpatched Critical LMCache Flaw Lets Unauthenticated Attackers Run Code Remotely
Unpatched LMCache flaw CVE-2026-105192 lets unauthenticated attackers run code via pickle over ZeroMQ.
76
autotrust/GLM5.3-Flash-E224-DGX-Spark — new model trending #29 on Hugging Face
autotrust released an unofficial pruned NVFP4 GLM-5.3-Flash build that fits two NVIDIA DGX Sparks.
44
Real Long-Term Memory for AI: A 50-Million-Token Window That Is Faster and Cheaper Than Recompute
Researchers store encrypted KV cache on NVMe so Gemma 4 recalls facts across 50 million tokens without recompute.
58
Can an Open Model Do Security Research? Cantina’s apex-flash-1 Solves 40 of 60 Held-Out Bug Tasks
Cantina and Yeta Labs released apex-flash-1, an open model that solved 40 of 60 bug tasks.
71
autotrust/GEV-26B-Decide-NVFP4 — new model trending #30 on Hugging Face
autotrust published GEV-26B-Decide-NVFP4, an NVFP4 expert-quantized 26B model that fits in about 17 GB.
48
nerkyor/Qwen3.8-27B-Coder390-EfficientThink-Opus5.5-GPT6Astra-Grok4.7-DSV4Pro-K3-SFT-RLOO-MTP-DFlash2 — new model trending #30 on Hugging Face
nerkyor released a Qwen3.8-27B post-train that cuts 94K truncations and lifts GPQA, MMLU, and coding scores.
42
Prime Intellect Launches Prime Inference: Serverless and Reserved Serving for Frontier Open Models
Prime Intellect launched Prime Inference, serverless and reserved serving for open models such as GLM-5.3.
52
Cohere Releases Embed 5: How It Compares to Voyage 4 Large, Gemini Embedding 2, and OpenAI
Cohere launched Embed 5 Pro and Fast, multimodal embeddings with a 128K context for search, RAG, and agents.
58
The vulnerabilities AI finds are the ones attackers want
GTIG says AI-found flaws skew toward remote code execution, and attackers hit BeyondTrust CVE-2026-1731 within four days.
80
Working Around the Compute Ceiling: Byte-Exact Memory in Galahad Makes LLM Reading a One-Time Cost LLM Reading a One-Time Cost
Galahad caches byte-exact KV state so repeated LLM document reading becomes a one-time cost.
52
SparseEngine: Sparse-First Inference Engine
SparseEngine, a sparse-first LLM inference engine, reports over 10x throughput with KV eviction versus existing serving stacks.
52
SlimWise: Decoupling Expert Pruning Across Prefill and Decode for Efficient MoE Serving
SlimWise prunes MoE experts only at decode, lifting Qwen3.6 throughput up to 1.81x in vLLM.
56
DeepEdu-v1: Efficient and Scalable Agentic LLMs for Vietnamese Education
DeepEdu-v1 tutors Vietnamese students on-premise, cutting retrieval calls 7.7x and lifting agentic accuracy from 70% to 79.5%.
41
EAServe: Encode-Aware Disaggregated Serving for Multimodal Large Language Models
EAServe raises multimodal LLM serving goodput by controlling encode, prefill placement, and GPU sharing.
48
Aikido Security Releases Altar-1: An Open-Weight Security Model Pruned From GLM-5.3 to 328 GB
Aikido released Altar-1, an open-weight security model pruned from GLM-5.3 to 328 GB.
56
Depth-adaptive Inference of Looped Language Models via Continuous Depth Batching
Continuous depth batching enables efficient depth-adaptive inference for looped language models, reaching up to 99% of estimated maximum speedup.
48
BottleCap AI Releases ThinkingCap-Qwen3.8-27B: 37.2% Fewer Thinking Tokens at a 0.86pp Accuracy Cost
BottleCap AI released ThinkingCap-Qwen3.8-27B, cutting thinking tokens 37.2% with a 0.86-point accuracy drop.
52
Contrastive-LM Releases CLM-8B: An Open System One Model That Scores Agent Actions Up to 9× Faster Than Jev
Contrastive-LM released CLM-8B, an open model that scores agent actions up to 9× faster than Jev.
58
orcarouter/OrcaSAQ-2-27B — new model trending #30 on Hugging Face
OrcaRouter released OrcaSAQ2 27B, a 12.3 GB 3-bit quant of Qwen3.8-27B with near-BF16 perplexity.
46
Nokia Open-Sources AnyJev: A Training-Free Layer That Turns Any Open LLM Into a Calibrated Decision Model
Nokia open-sources AnyJev, a training-free library that converts any open LLM into a calibrated decision model for classification tasks.
50
Measuring the Serving Stack Instead of the Model: Hidden Confounds in Local Tool-Use Evaluation
Local serving stacks can distort tool-use scores, including false 0% fidelity for rejected models.
43
Aikido Security Unveils Altar-1 Open-Weight AI for Cybersecurity Defense
Aikido released Altar-1, an open-weight defensive model pruned from GLM-5.3 for on-premises security testing.
54
Related CVEs
by mentions- CVE-2026-4227191%8.7Command Injection in BerriAI LiteLLM AI Gateway Exploited in the Wild
- CVE-2026-1051921%9.8Unauthenticated pickle RCE in LMCache distributed mode
- CVE-2026-173193%9.9Pre-Authentication OS Command Injection RCE in BeyondTrust Remote Support and PRA
- CVE-2025-3248100%9.8Unauthenticated RCE in Langflow /api/v1/validate/code
- CVE-2026-077063%9.8Unauthenticated Remote Code Execution in Langflow validate Endpoint
- CVE-2024-58286<1%9.3dizqueTV 1.5.3 contains a remote code execution vulnerability that allows attackers to inject arbitrary commands through the FFMPEG Executable Path settings
- CVE-2026-502736%8.8Authenticated Path Traversal Arbitrary File Write in Langflow Files API
- CVE-2026-43503<1%8.8Linux kernel: net: skbuff: propagate shared-frag marker through frag-transfer helpers Two frag-transfer helpers…
- DoS via unvalidated cache_salt in vLLM before 0.30.0
- File and Print Sharing service in Windows 95, Windows 98, and Windows Me does not properly check the password for a file share, which allows remote attackers…
Appears with
- NVIDIAvendor · 12
- SGLangproduct · 12
- Hugging Faceorganization · 10
- llama.cppproduct · 8
- Googlevendor · 7
- GLM-5.3ai model · 5
- Ollamaproduct · 5
- NVIDIAorganization · 4
- Qwen3.8-27Bai model · 4
- MLXproduct · 4
- Gemini 4 Argonai model · 3
- Gemma 4 E2Bai model · 3
- Z.aivendor · 3
- Transformersproduct · 3
- autotrustorganization · 2
- LiteLLMproduct · 2
Entities are extracted by the model from each article. First seen , last mentioned . Watching keeps it in this browser only (no account); the watchlist page and dashboard alerts use it.