Public story · 2026-03-22 · source-backed

Mistral Small 4: 119B MoE Unifies Instruct, Reasoning, Vision, and Coding

Issue

2026-03-22

Confidence

source-backed

Sources

Redaction

passed

Story

6B active parameters per token, 256K context window, consolidates four previously separate models into one deployment target. 40% faster, 3x higher throughput than Small 3. Apache 2.0 on Hugging Face. The most capable open-weight model requiring no model switching across multimodal and agentic workflows. Source

Hugging Face partners with Open Secure AI Alliance / Shared entity: Apache / Same source domain / Shared topic / What happened next
Mistral ships a safety model that takes plain-English rules
Linked by a graph relationship (Hugging Face partners with Open Secure AI Alliance); both cover Apache; reported by the same outlet (mistral.ai).
Mistral released Mistral Small / Shared entities / Same source domain / Shared topic / What happened next
Devstral 2 is open source, 123B params, hits 72.2% SWE-bench, and costs a seventh of Sonnet
Linked by a graph relationship (Mistral released Mistral Small); both cover Apache, Small; reported by the same outlet (mistral.ai).
Hugging Face partners with Open Secure AI Alliance / Shared entity: Hugging Face / Shared topic / What happened next
NVIDIA forms security alliance, leaves out OpenAI and Anthropic
Linked by a graph relationship (Hugging Face partners with Open Secure AI Alliance); both cover Hugging Face; overlapping topics (face, hugging, model).
Hugging Face partners with NVIDIA / Shared entity: Hugging Face / Shared topic / What happened next
Qwen jumped from 1% to 69% of open-model fine-tunes
Linked by a graph relationship (Hugging Face partners with NVIDIA); both cover Hugging Face; overlapping topics (face, hugging, model).
Hugging Face partners with NVIDIA / Shared entities / Shared topic / What happened next
Meta ships Muse Glimmer, a 30B open agent model
Linked by a graph relationship (Hugging Face partners with NVIDIA); both cover Apache, Hugging Face; overlapping topics (apache, model).
Mistral released Mistral Small / Shared entities / Shared topic / What happened next
Thinking Machines Ships Inkling: 975B Open Weights, and It's Behind Two Chinese Labs
Linked by a graph relationship (Mistral released Mistral Small); both cover Apache, Hugging Face, Small; overlapping topics (active, agentic, context).
Hugging Face partners with NVIDIA / Shared entities / Shared topic / What happened next
DeepSeek V4 Preview Drops: 1.6T Parameters, 1M Context, Apache 2.0, and a 35x Price Gap That Should Make Every Builder Rethink Their Inference Budget
Linked by a graph relationship (Hugging Face partners with NVIDIA); both cover Apache, Hugging Face; overlapping topics (apache, context).
Mistral released Mistral Small / Shared entity: Apache / Same source domain / Shared topic / Earlier coverage
Mistral 3 Family + Devstral 2 + Vibe CLI — Apache 2.0 Open-Source Coding Agent
Linked by a graph relationship (Mistral released Mistral Small); both cover Apache; reported by the same outlet (mistral.ai).

Source trail

Source

Entities

Apache Coding Hugging Face Mistral Small MoE Unifies Instruct Reasoning Small Vision

Provenance

Canonical issue: Ramsay Research Agent — 2026-03-22
AI generated: no
Story unit: 2026-03-22-mistral-small-4-119b-moe-unifies-instruct-reasoning-vision-and-coding
Labels: source-backed, canonical briefing excerpt

Fetching from the wire…

Mistral Small 4: 119B MoE Unifies Instruct, Reasoning, Vision, and Coding

Story

Related stories

Mistral ships a safety model that takes plain-English rules

Devstral 2 is open source, 123B params, hits 72.2% SWE-bench, and costs a seventh of Sonnet

NVIDIA forms security alliance, leaves out OpenAI and Anthropic

Qwen jumped from 1% to 69% of open-model fine-tunes

Meta ships Muse Glimmer, a 30B open agent model

Thinking Machines Ships Inkling: 975B Open Weights, and It's Behind Two Chinese Labs

DeepSeek V4 Preview Drops: 1.6T Parameters, 1M Context, Apache 2.0, and a 35x Price Gap That Should Make Every Builder Rethink Their Inference Budget

Mistral 3 Family + Devstral 2 + Vibe CLI — Apache 2.0 Open-Source Coding Agent

Source trail

Entities

Provenance

Mistral Small 4: 119B MoE Unifies Instruct, Reasoning, Vision, and Coding

Story

Related stories

Mistral ships a safety model that takes plain-English rules

Devstral 2 is open source, 123B params, hits 72.2% SWE-bench, and costs a seventh of Sonnet

NVIDIA forms security alliance, leaves out OpenAI and Anthropic

Qwen jumped from 1% to 69% of open-model fine-tunes

Meta ships Muse Glimmer, a 30B open agent model

Thinking Machines Ships Inkling: 975B Open Weights, and It's Behind Two Chinese Labs

DeepSeek V4 Preview Drops: 1.6T Parameters, 1M Context, Apache 2.0, and a 35x Price Gap That Should Make Every Builder Rethink Their Inference Budget

Mistral 3 Family + Devstral 2 + Vibe CLI — Apache 2.0 Open-Source Coding Agent

Source trail

Entities

Provenance