Cohere’s Command A+ ships low-compute agentic AI

// 2h agoMODEL RELEASE

Cohere’s Command A+ ships low-compute agentic AI

Command A+ is Cohere’s latest language model release: an Apache 2.0, mixture-of-experts enterprise workhorse built for reasoning, tool use, multimodal inputs, and multilingual workflows. Cohere positions it as its fastest and most powerful model yet, with deployment targets as light as two H100s or a single B200, making it a strong fit for teams that want high-throughput agentic capability without the usual infrastructure tax.

// ANALYSIS

This is a pragmatic enterprise release, not a vanity benchmark drop: Cohere is pairing solid capability with a licensing and deployment story that enterprise buyers can actually use.

–The Apache 2.0 license lowers friction for real-world adoption, especially for private and regulated deployments.
–The MoE design and low-bit quantization targets are the real differentiator here: it is built for efficiency, not just leaderboard optics.
–Multimodal, multilingual, and tool-use support make it more broadly useful than a narrow reasoning-only model.
–The launch reads as a consolidation move across the Command family, which should simplify product and infra choices for customers.

// TAGS

open-sourceenterprise-aillmagentmultimodalmultilingualcoheremoe

DISCOVERED

2h ago

2026-05-23

PUBLISHED

7h ago

2026-05-23

RELEVANCE

9/ 10

// KEEP READING

More AI developer news from the feed

EXPLORE FULL FEED

RESEARCH16m ago

Life-Harness boosts frozen LLMs 88% without fine-tuning

Researchers at Peking University introduce Life-Harness, an open-source runtime framework that improves frozen LLM agent performance by 88.5% on average. Instead of fine-tuning model weights, it modifies the interaction interface to dynamically correct formatting errors, enforce environment rules, and prevent trajectory failure loops.

RESEARCH1h ago

Nous Research drops Contrastive Neuron Attribution method

Nous Research has unveiled Contrastive Neuron Attribution (CNA), a mechanistic interpretability method that steers LLM behavior by isolating sparse circuits of under 200 neurons. The technique enables precise suppression or amplification of specific behaviors like refusal without degrading model coherence.

TUTORIAL3h ago

Anthropic engineer details internal Claude Code loop

Anthropic engineer Thariq Shihipar details the "Prompt, Plan, Verify" loop used internally to build Claude Code. The workflow replaces traditional autocomplete with an agentic cycle of architectural planning and empirical verification.