The Open Weights
LatestModelsLeaderboardsCompanies
Subscribe
The Open Weights

The daily record of open-source AI. New model releases, leaderboards, and what's coming next — written for people who ship.

Refreshed every 12 hours

Discover

  • Latest releases
  • New today
  • Trending models

Browse

  • All models
  • Companies
  • Categories
  • Leaderboards

About

  • About
  • Editorial policy
  • Hardware estimates
  • RSS feed
  • llms.txt
  • Newsletter

© 2026 The Open Weights. An independent publication.

PrivacyTermsSMSAggregated by Claude · curated by humans.
LatestMistral AI4
Mistral AIText / LLM

Mistral Large 4 arrives as a trillion-param MoE

The new flagship is a sparse mixture-of-experts model with roughly 49B active parameters per token.

Oct 6, 2026
Major releaseOther

Mistral AI has introduced Mistral Large 4, the latest flagship in its Large series and the company's biggest model to date. According to the announcement, it is a mixture-of-experts (MoE) design with around one trillion total parameters but only about 49 billion active per token — a structure that aims to deliver the capacity of a very large model while keeping inference costs closer to a mid-sized one.

The model is positioned as a text and reasoning system, placing it among the growing class of frontier models tuned for multi-step problem solving rather than raw chat alone. Mistral is framing this release as a preview, so some details and availability specifics are still settling.

Why it matters

Sparse MoE architectures have become the dominant approach at the high end, and the 49B-active figure is the number worth watching. It signals that Mistral wants Large 4 to be deployable at a reasonable serving cost despite its trillion-parameter footprint.

  • Scale: roughly 1T total parameters, a step up for the Large family
  • Efficiency: about 49B active parameters per token via MoE routing
  • Focus: text generation plus reasoning

Licensing terms for Large 4 fall outside the standard open templates, so teams evaluating it should read the usage conditions closely before committing. As Simon Willison noted in his writeup, the headline here is sheer size — but the practical story will depend on how the active-parameter economics hold up in real deployments.

Sources

  • Introducing Mistral Large 4: Le chonk

    Announcement

    Visit
OlderMiniCPM-V 4.7 brings a 35B MoE vision model on-deviceOpenBMB · Vision-Language · 2 months agoNewerOpenBMB's MiniCPM-V 4.7 brings MoE to vision modelsOpenBMB · Vision-Language · 2 months ago

Get the model

Announcement

Specs

Parameters1000B · MoE
Active params49B active
LicenseOTHER

Can you run it?

Datacenter only — about 603.8 GB even at 4-bit.

  • BF16 (as published)

    Multi-node cluster

    2.0 TB
  • 8-bit

    Multi-node cluster

    1.1 TB
  • 4-bit

    8×H100 80 GB node

    603.8 GB
Your machine
  • Mixture-of-experts: all 1000B parameters must be resident even though only 49B are active per token — active count affects speed, not memory.
  • Precision not published; assuming 16-bit (BF16) weights.
  • Parameter count is an editorial estimate — no weights metadata on Hugging Face.
  • KV cache approximated from model size (no config metadata) at 8,192 tokens.

Estimated from the parameter count. Assumes an 8K context and ~1 GB runtime overhead; actual needs vary. How we estimate


Modalities

Text / LLMReasoning

The Weekly Weights

Every open release that mattered, one email a week.

0 comments

No comments yet. Be the first to weigh in.

More from Mistral AI

All Mistral AI releases →
Mistral Large 4 (1T-A52B)
Mistral AI/Text / LLMDatacenter

Mistral Large 4 Arrives as a 1T-Parameter MoE

The new flagship pairs a trillion total parameters with 52B active per token and adds vision alongside multilingual text.

Oct 29, 2026
Timestral-209M
Mistral AI/Text / LLMRuns anywhere

Mistral's Timestral brings forecasting to open weights

The French lab releases a compact 209M-parameter foundation model aimed squarely at time-series prediction.

Oct 14, 2026
Mistral AI/Vision-LanguageRuns anywhere

Mistral debuts Shieldstral, a 3B safety model

The open-weights multimodal moderation model brings content safety checks to both text and images under an Apache 2.0 license.

Aug 4, 2026

More in Text / LLM

All Text / LLM →
Aion-1.0-Instruct-2.3B
Microsoft/Text / LLMRuns anywhere

Microsoft's Aion 1.0 Instruct is a compact 2.3B chat model

A small, multilingual instruction-tuned model aimed at chat and summarization lands on Hugging Face.

Nov 18, 2026
SenseNova-Loop-9B
SenseTime/Vision-LanguageRuns on a laptop

SenseTime Releases Looped 9B Multimodal Model

The company's new SenseNova-Loop-9B pairs vision-language capabilities with a recurrent looped-transformer design.

Nov 17, 2026
Hunyuan Hy4-flash
Tencent/Vision-LanguageMulti-GPU server

Tencent releases Hunyuan Hy4-flash multimodal MoE

The lightweight, mixture-of-experts model adds vision and text instruction-following to Tencent's open Hunyuan lineup.

Nov 17, 2026