Public story · 2026-03-17 · source-backed

Mistral Small 4: 119B MoE, 6B Active, Apache 2.0.

Issue

2026-03-17

Confidence

source-backed

Sources

Redaction

redacted

Story

A reasoning_effort parameter (none/high) gives per-request cost-vs-depth control. 40% faster, 3x more requests/second than predecessor. First single checkpoint unifying vision, code agents, and configurable CoT under a permissive license. Day-0 on NVIDIA NIM, vLLM, llama.cpp, HuggingFace. Source

HuggingFace released Claude Code / Shared entities / Shared topic / Earlier coverage / Tension
Run Claude Code Against GLM-5 (744B) for Free via NIM Proxy
Linked by a graph relationship (HuggingFace released Claude Code); both cover MoE, NVIDIA NIM; overlapping topics (active, code, license).
HuggingFace released Claude Code / Shared entities / Shared topic / What happened next
Local models became a first-class feature of hosted coding agents in a 48-hour window
Linked by a graph relationship (HuggingFace released Claude Code); both cover Apache, MoE; overlapping topics (agent, code, llama).
Mistral released Mistral Small / Shared entities / Shared topic / What happened next
Google Ships Gemma 4 Under Apache 2.0. Four Models, Arena #3, and the License That Changes Everything.
Linked by a graph relationship (Mistral released Mistral Small); both cover Apache, HuggingFace, MoE; overlapping topics (apache, license, llama).
HuggingFace released Claude Code / Shared entity: Apache / Shared topic / What happened next / Tension
Claude Code Hits 44.4% of Hugging Face Agent Traffic
Linked by a graph relationship (HuggingFace released Claude Code); both cover Apache; overlapping topics (agent, apache, code).
HuggingFace released Claude Code / Shared entity: Apache / Shared topic / Earlier coverage
Rivet Sandbox Agent SDK — Universal API for Coding Agents (960 stars)
Linked by a graph relationship (HuggingFace released Claude Code); both cover Apache; overlapping topics (agent, apache, code, control).
HuggingFace released Claude Code / Shared entity: Apache / Shared topic / What happened next / Tension
agentmemory puts a dollar figure on memory architecture: ~$10/year vs ~$500/year.
Linked by a graph relationship (HuggingFace released Claude Code); both cover Apache; overlapping topics (agent, code).
HuggingFace released Claude Code / Shared entity: MoE / Same source domain / What happened next / Tension
A 27B Model Just Matched the 400B One. It Fits on Your Laptop.
Linked by a graph relationship (HuggingFace released Claude Code); both cover MoE; reported by the same outlet (simonwillison.net).
HuggingFace released Claude Code / Shared entities / Shared topic
GLM-5 as Claude Code Alternative via NVIDIA NIM Free Tier.
Linked by a graph relationship (HuggingFace released Claude Code); both cover MoE, NVIDIA NIM; overlapping topics (active, code).

Source trail

Source

Entities

Active Apache CoT HuggingFace Mistral Small MoE NVIDIA NIM

Provenance

Canonical issue: Ramsay Research Agent — 2026-03-17
AI generated: no
Story unit: 2026-03-17-mistral-small-4-119b-moe-6b-active-apache-2-0
Labels: source-backed, canonical briefing excerpt

Fetching from the wire…

Mistral Small 4: 119B MoE, 6B Active, Apache 2.0.

Story

Related stories

Run Claude Code Against GLM-5 (744B) for Free via NIM Proxy

Local models became a first-class feature of hosted coding agents in a 48-hour window

Google Ships Gemma 4 Under Apache 2.0. Four Models, Arena #3, and the License That Changes Everything.

Claude Code Hits 44.4% of Hugging Face Agent Traffic

Rivet Sandbox Agent SDK — Universal API for Coding Agents (960 stars)

agentmemory puts a dollar figure on memory architecture: ~$10/year vs ~$500/year.

A 27B Model Just Matched the 400B One. It Fits on Your Laptop.

GLM-5 as Claude Code Alternative via NVIDIA NIM Free Tier.

Source trail

Entities

Provenance

Mistral Small 4: 119B MoE, 6B Active, Apache 2.0.

Story

Related stories

Run Claude Code Against GLM-5 (744B) for Free via NIM Proxy

Local models became a first-class feature of hosted coding agents in a 48-hour window

Google Ships Gemma 4 Under Apache 2.0. Four Models, Arena #3, and the License That Changes Everything.

Claude Code Hits 44.4% of Hugging Face Agent Traffic

Rivet Sandbox Agent SDK — Universal API for Coding Agents (960 stars)

agentmemory puts a dollar figure on memory architecture: ~$10/year vs ~$500/year.

A 27B Model Just Matched the 400B One. It Fits on Your Laptop.

GLM-5 as Claude Code Alternative via NVIDIA NIM Free Tier.

Source trail

Entities

Provenance