AI News

Google introduces Nano Banana 2.1 image generation and editing model

Google has announced Nano Banana 2.1, an upgraded image generation and editing model that reports improvements in visual design, mask-based editing, and subject consistency.

Replit adds cross-project context to chat for referencing files and background edits

Replit now supports context across all user projects within chat. Users can prompt Replit to locate existing projects, use one project as a reference to modify another, and execute edits as background tasks.

Splunk expands work with awscloud and NVIDIA to bring local compute to Splunk AI

At splunkconf26, Splunk announced expanded collaboration with awscloud and NVIDIA to deliver more local compute power for Splunk AI, enabling customers to run governed AI on sensitive machine data in self-managed environments.

LangChain adds free p0-powered web search to Managed Deep Agents

LangChain announced that developers can add web search powered by p0 to Managed Deep Agents for free, without separate vendor accounts, extra key management, or search tool wiring.

Meta AI introduces MIRA architecture for long-horizon research agents

Meta AI researchers introduced MIRA, a dual-component agent architecture that decouples meta-reasoning from execution to improve decision-making in long-horizon research tasks.

OpenRouter lists Mistral Large 4 preview with 1M context and $0.07 cached input

Mistral Large 4 is in public preview on OpenRouter with a 1M context window and cached input at $0.07 per million tokens. For the first two weeks, the 50% discount also sets input at $0.68 and output at $2.09 per million tokens.

Mistral Large 4 is now available in Arena.ai's Agent Arena and Code Arena

Arena.ai has added Mistral AI's Mistral Large 4 to Agent Arena for user testing alongside Code Arena across WebDev, Text, and Vision. The model features 1 trillion total parameters with 49 billion active parameters, with evaluation scores coming soon.

Command Code launches ACP support for compatible editors like Zed

Command Code has added ACP support, allowing developers to run Command Code workflows in ACP-compatible editors such as Zed with consistent tools, skills, permissions, and sessions.

OpenRouter launches Mistral Large 4 in public preview with 50% launch discount

Mistral Large 4 is now available on OpenRouter in public preview with a 512K context window and up to 256K output tokens. OpenRouter is offering a 50% discount for the first two weeks at $0.68 per million input tokens and $2.09 per million output tokens.

LangChain outlines Rippling AI architecture built on Deep Agents and LangSmith

Rippling AI is built on Deep Agents and LangSmith, employing a supervisor agent to coordinate five to seven specialized subagents alongside LangSmith tracing, evaluations, and production monitoring.

Mistral Large 4 evaluated by Artificial Analysis with 512k context and API pricing

Artificial Analysis evaluated Mistral Large 4 Preview, which scored 38 on the Intelligence Index and 50% on the Cyber Index. Mistral AI is offering the 512k-context model at $1.36 per 1M input tokens and $4.18 per 1M output tokens, with a 50% discount during its first two weeks.

Lambda adds NVIDIA AI AIPerf benchmarks to Model Cards

Lambda updated its model cards to include serving benchmarks measured with NVIDIA AI's AIPerf package on minimal Lambda hardware, showing deployment behavior across user concurrency thresholds.

Victor M says Mistral 4 is coming to Hugging Face soon

Victor M says Mistral 4 will be on Hugging Face soon and links to the Mistral-Large-4.0-1T05-A52B model page. The note does not give a release date or say the weights are available now.

NVIDIA AI says ATT, SoftBank, and Indosat are adopting open models for telecom agents

ATT, SoftBank, and Indosat are deploying open models for autonomous network operations and local language adaptation using a full-stack framework provided by NVIDIA AI.

Mistral AI unveils Mistral Large 4 with 1T parameters and immediate API access

Mistral AI announced Mistral Large 4, a 1-trillion parameter natively multimodal model with 49 billion active parameters. The model is available now via API and Mistral Cloud infrastructure, ahead of an open weights release planned for late October.

Ant Group releases Ling 3.1 Flash, scoring 41 on Artificial Analysis Intelligence Index

Ant Group has released Ling 3.1 Flash, a reasoning model featuring 560 billion total parameters and a 1 million token context window. Artificial Analysis evaluated the model at 41 on its Intelligence Index v4.3, highlighting notable improvements in agentic performance and token efficiency.