LLM News

Every LLM release, update, and milestone.

0
model releaseGoogle

Google Antigravity adds Gemini 4 Hydrogen and Orchestrator, which switches between Hydrogen and Carbon mid-session

Google is rolling out Gemini 4 Hydrogen in Antigravity, along with Orchestrator, a model option that hands work back and forth between Hydrogen and Gemini 4 Carbon within one session. Google claims Orchestrator used 24% less quota and finished 19% faster than Carbon on Terminal-Bench 4.1 while reaching near-Carbon task success.

0
researchIBM

IBM releases Granite 4.1 3B and 8B LLMON previews to separate instructions from data

IBM released two experimental preview models, granite-4.1-3b-LLMON-preview and granite-4.1-8b-LLMON-preview, trained on LLMON, an XML-like markup that uses reserved special tokens to mark structure. IBM claims the format nearly eliminates indirect prompt injection on some benchmarks while keeping general capability close to Granite 4.1 Instruct.

3 min readvia huggingface.co ↗
0
product updateAnthropic

Anthropic Launches Claude for Manufacturing, a 17-Skill Plugin for Quoting, Quality and Maintenance

Anthropic launched Claude for Manufacturing on November 12, 2026. The plugin has 17 skills for quoting, quality investigations, production reporting, maintenance diagnosis, sourcing and design. It works from a manufacturer's existing records and sends processes to employee review.

1 min readvia unite.ai ↗
0
model releaseGoogle

Google launches Gemini 4 Hydrogen: 1M-token context, $1.50 input and $7.50 output per 1M tokens

Google announced Gemini 4 Hydrogen on November 12, 2026, with native 1 million-token input and up to 1 million output tokens. It costs $1.50 per million input tokens and $7.50 per million output tokens, and Google claims it scores 51 on the Artificial Analysis Intelligence Index at its default high thinking level.

3 min readvia 9to5google.com ↗
0
analysisAnthropic

VulnCheck: Project Glasswing has 37,966 candidate vulnerabilities, 217 marked fixed in public records

Security firm VulnCheck reports that seven months after launch, Anthropic's Project Glasswing has produced 37,966 candidate vulnerabilities, with 217 marked fixed in public disclosure records. Anthropic's own figures, as of November 2, show 629 upstream patches, and both sides point to human review as the bottleneck.

0
model release

Baidu launches ERNIE Lightning reasoning model and ERNIE-Wizshot-ME video model on Qianfan

Baidu released ERNIE Lightning, a deep-thinking model built first for Baidu Search, and ERNIE-Wizshot-ME, an expressive video model, at Baidu World 2026 on November 12. Both are live on Baidu AI Cloud's Qianfan platform with APIs open to enterprise customers and partners. Baidu has not published benchmarks, parameter counts or context lengths for either.

3 min readvia pandaily.com ↗
0
model releaseByteDance

ByteDance launches Seedance 2.5 Flash video model at about 0.99 yuan per second for 720p, 34% below Seedance 2.5

ByteDance's Volcano Engine released Seedance 2.5 Flash on November 12, pricing 720p text-to-video and image-to-video at about 0.99 yuan per second, roughly 34% below Seedance 2.5. The company claims it generates video twice as fast as Seedance 2.5. API access on Volcano Ark opens November 15.

0
product updateTencent

Tencent launches Yuanbao Search, a standalone AI search product at so.qq.com

Tencent Cloud announced Yuanbao Search on November 11, a standalone AI search product reachable at so.qq.com and selectable as a search engine in the latest QQ Browser. Tencent says it rebuilds the full search pipeline on the large-model capabilities behind its Yuanbao assistant and draws on an index of hundreds of billions of pages.

3 min readvia pandaily.com ↗