LoRA merge audit finds bad adapters inflate forgetting scores
Naive merging of an audited four-adapter set cut GSM8K by 10.5 points and HumanEval by 18.3.
Stories per day · 30 days
3 RSS
Relatedopen-weights7forecasting3llama3mistral-large-43speech-to-text3text-to-speech3time-series3timestral3voice-agents3voxtral3coding-agents2hugging-face2model-release2nvidia2sovereign-ai2svd2agentic-ai1ai-misuse1
30 stories, newest first · one card per story
Stories tagged #mistral
Mistral open-sources Large 4 weights despite cyberattack uplift warning
Downloaded weights sit outside API guardrail updates, and Mistral has not said whether Large 4 meets the EU systemic-risk threshold.
1
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.
4
Mistral releases Large 4 open weights, nicknamed Le Chonk
The trillion-parameter model leads some open coding scores but trails top Chinese models and Claude overall.
1
Mistral releases open weights for Mistral Large 4
The 1-trillion-parameter multimodal model adds three reasoning modes and reported defensive-security gains.
3
Mistral releases open weights for trillion-parameter Mistral Large 4
Nicknamed Le Chonk, ML4 is on Hugging Face and Mistral's API, and the company claims 47 percent fewer tokens than GLM-5.3.
1
Mistral Large 4, nicknamed Le Chonk, on Hugging Face
The post announces the model on Hugging Face; no license, weights, or benchmark details are in the captured item.
Component translation does not reliably change LLM refusal rates
On 190 harmful behaviours, only an exploratory Mistral Small 4 Spanish swap stayed significant after correction.
LEMS and KFAC-SVD improve SVD compression of large language models
Zero-shot gains reach 4.8 points through 70B models, and KFAC-SVD lowers perplexity 15% on average.
LEMS advances SVD compression for LLMs
LEMS and KFAC-SVD deliver up to 4.8 percentage point accuracy gains on Mistral, Qwen3 and Llama 3 models.
Mistral and Nvidia partner on physics AI for industrial design
They plan to link Mistral Large 4 and physics models with CUDA-X, Warp and Omniverse, plus public benchmarks and datasets.
Mistral AI releases Voxtral 3 speech models and voice SDK
The captured page only preserves the release title; sizes, benchmarks, and availability are not stated.
Mistral launches Voxtral 3 real-time voice agent stack
The stack pairs transcription, speech synthesis and a LiveKit SDK, priced from $0.0035 per minute.
Mistral releases Voxtral 3 speech models and voice agent SDK
The family adds multilingual transcription and speech synthesis, offered on an EU-hosted API or for self-hosting.
CERN and Mistral plan a foundation model for collider physics
The planned Large Physics Model would train on billions of simulated LHC collisions and is not released.
Mistral releases Timestral time-series foundation models
Mistral says its 209M model ranks first on GIFT-Eval, TIME and fev-bench while using about two-thirds the parameters of the prior leader.
Mistral releases Timestral, a 209M forecasting model family
The title says the family outperforms Google’s TimesFM-3; the article body was not in the provided text.
Mistral releases Timestral-209M time-series forecasting model
The 209M checkpoint forecasts up to 1,024 steps from 32,768 history steps across 21 quantiles.
From Earth to Orbit, France’s Ecosystem Delivers Agentic Computing at Scale With NVIDIA
NVIDIA highlights French firms deploying agentic AI, from an orbital Jetson satellite to Nemotron tools and Mistral Large 4.
[OSSA-2026-044] OpenStack Mistral: Four authorization and privilege vulnerabilities in Mistral (CVE-2026-93858, CVE-2026-93860, CVE-2026-93861, CVE-2026-97147))
OpenStack Mistral has four authorization and privilege flaws in OSSA-2026-044.
Mistral says "Le Chonk" can challenge the best AI models
Mistral released Le Chonk, a 1-trillion-parameter open-weight model it says rivals leading US and Chinese systems.
llm-mistral 0.16
llm-mistral 0.16 adds LLM CLI support for Mistral reasoning models, including Mistral Large 4.
Mistral Large 4: "Le Chonk"
Mistral previews Mistral Large 4, a 1-trillion-parameter open-weight multimodal model, with weights due by month's end.
9
Align Then Reason: A Multimodal Lip-Sync Judge for Dubbing
Align Then Reason judges multilingual dubbing lip-sync by aligning lip frames to phonetics before LLM reasoning.