AI & LLMs
Claude Duet is Anthropic's proactive agent app, in preview
Anthropic opened a waitlist preview of Claude Duet on October 21, 2026: an Opus 5.5 agent that works from its own cloud computer and messages you first.
AI & LLMs - Page 2
AI models, agents, and research - from a developer's point of view.
AI & LLMs
OpenAI GPT-6 models come to Google Cloud in preview
Google Cloud began a private preview of OpenAI's GPT-6 Astra, GPT-6.1 Sol and GPT-6 Luna on October 21, 2026, billed through its Marketplace.
AI & LLMs
ElevenLabs Eleven Realtime voice agents reply in 570 ms
ElevenLabs' Eleven Realtime v1 Max answers in 570 ms and scores 98.8% on Big Bench Audio, per the company. Its API copies OpenAI's Realtime format.
AI & LLMs
Inception Mercury 3 diffusion LLM claims 1,068 tokens/sec
Inception's Mercury 3 diffusion LLM claims 1,068 tokens a second, a 1M-token context and $0.20/$0.70 per million tokens, half off at launch.
AI & LLMs
Mistral Voxtral 3 adds voice agents from $0.0035 a minute
Mistral's Voxtral 3 adds speech-to-text from $0.0035 a minute, a $13 per million character voice model and a voice agent API from $0.04 a minute.
AI & LLMs
MiniMax M3.1 Flash rolls out, claims 70% cheaper agents
MiniMax's new model costs $0.15 per million input tokens and $0.60 output, and the company says it runs AI agents for 70% less than M3.
AI & LLMs
AMD GAIA 0.25 folds its local AI agents into one
AMD's open-source local agent app now runs one agent, shrinks its Windows Python setup from 0.9GB to 0.1GB and stops crashing on 16GB laptops.
AI & LLMs
Cloudflare Clairic model returns typed values, not text
In Cloudflare's own claims test, a pipeline led by its new Clairic model ran in 19.3 seconds instead of 61.7 and cost 49% less than an LLM alone.
AI & LLMs
Claude Opus 4.7 agents invent games on a virtual island
An MIT study left 13 Claude Opus 4.7 agents alone on a virtual island for 30 hours. They invented games, then beat fresh copies on new tasks.
AI & LLMs
Schwab, JPMorgan block AI agents from moving client money
Schwab is blocking outside AI agents such as Meta's Muse from moving client money and will offer read-only MCP access instead. JPMorgan blocks them too.
AI & LLMs
Reflection Beam open weights land under Apache 2.0
Reflection put Beam's 501B-parameter weights on Hugging Face on October 16, 2026 under Apache 2.0. The FP8 and 4-bit versions fit on one 8-GPU server.
AI & LLMs
StepFun Step 5 Preview: 600B open-weight agent model
StepFun's Step 5 Preview, a 600B-parameter model with 27B active and a 1M-token context, is on Hugging Face after an API-only month at $1/$2.70 per M.
AI & LLMs
OpenAI reports 4 more model incidents, 2 rated critical
OpenAI posted four new misalignment reports on October 15, 2026. In one, a model forged tokens to save 24 years of data from a government website.
AI & LLMs
Elorian Glimpse model rechecks images while it reasons
Elorian's first model, Glimpse, costs $0.60 per million input tokens and looks back at images as it reasons. It beat GPT-6 Astra on 4 of 11 tests.
AI & LLMs
Ling-3.1-flash: Ant Group opens 560B model under MIT
Ant Group's inclusionAI published Ling-3.1-flash weights under the MIT license: 560B parameters, 25B active per token and a 1M-token context.
AI & LLMs
DeepSeek V4.1-Flash-1017 raises agent scores, keeps price
DeepSeek's V4.1-Flash-1017 update scores 77.0% on DeepSWE, up from 74.2%, at unchanged API prices, and adds Search and Fetch APIs at $0.75 per 1,000 calls.
AI & LLMs
CERT-In says AI agents under test reached Indian systems
India's CERT-In says AI agents under test used exposed credentials to reach live systems at Indian firms, in a critical advisory issued October 15, 2026.
AI & LLMs
FLUX 3 Image Klein ships as a 5.6B open-weight model
Black Forest Labs released FLUX 3 Image Klein, a 5.6B open-weight image model that makes and edits pictures, with free commercial use under $5M revenue.
AI & LLMs
Microsoft MAI-Base-2 completes pre-training at 2.8T params
Microsoft AI finished pre-training MAI-Base-2, a 2.8T-parameter model with 80B active, on 16,000 GB200 GPUs in 38 days. It is not released yet.
AI & LLMs
Nvidia Nemotron 3.5 Super VL ships open 120B weights
Nvidia's Nemotron 3.5 Super VL is an open-weight model with 120B parameters, 12B active, that reads text, images and video under the OpenMDW license.
AI & LLMs
Jais 3 brings six open Arabic models, up to 375B
MBZUAI's IFM and Inception42 released Jais 3 on October 15: six open Arabic models from 0.9B to 375B parameters, plus Jais Speech, under Apache 2.0.
AI & LLMs
Bank of America puts 1,000 contractors on AI-found flaws
Bank of America brought in 1,000 contractors to fix software flaws that AI models, including Anthropic's Mythos, found, CEO Brian Moynihan said on October 14.
AI & LLMs
Anthropic gives six METR evaluators employee-level access
Anthropic gave six METR researchers employee-level access to its systems and gave up its right to block what they publish about its risks and incidents.
AI & LLMs
xAI Grok Voice 3.0 ships at $0.08 a minute
xAI's Grok Voice 3.0 speech model costs $0.08 a minute in its Voice Agent API, the same as 2.0, and scores 86.9 on Artificial Analysis's speech index.
The weekly digest
One email every Friday with the week's top stories from all six desks: AI, dev tools, coding, the tech industry, startups and what's next. Free, no spam.