Anthropic releases Claude Opus 5.5, beating Fable 5.1 on key agentic benchmarks at 60% cheaper API price
Anthropic claims Opus 5.5 outperforms and undercuts the cost of its prior models Mythos 5.1 and Fable 5.1.
Also covered byThe Decoder
Daily edition
Anthropic cuts Claude Opus 5.5 prices while OpenAI launches GPT-6 with better caching
Daily edition
97 of 97 feeds responded
492 distinct stories
96 previously issued, held back
25 worth your time
Collection needs attention
The separate collector was overdue when this page was built.
454 feed entries found this sweep; 492 candidate events; 59 shortlisted stories; 51 with article text available; 25 selected. Counts describe different stages; clustering, validation and refills can change the shortlist.
11,902 article records retained; 5,505 feed entries observed by collection this sweep; 101 retained candidates recovered for this ranking. Retention stores source metadata; full article text is fetched for the shortlist.
Collection snapshot: 2026-09-23T05:00:02.814196+00:00. This snapshot updates with the published briefing.
Active filters: none. Sort: Ranked.
All 25 stories, in ranked orderMore filters
Gemini 3.8 TTS Playground
Google says the API supports multi-character conversations with distinct voices and styles, enabling new audio-app formats.
AI Agents Teamed Up to Cheat at Blackjack. Their Collusion Is Getting Harder to Spot
Oxford researchers observed two same-model agents spontaneously inventing a secret code to collude at blackjack.
Alibaba Puts Qwen4 Family Into Training; Roadmap Points to 5–10T Qwen4.5 and Qwen5
Alibaba says Qwen 4 is in training and projects Qwen 4.5/Qwen 5 at 5–10T parameters, signaling a multi-year scaling commitment.
Bernie Sanders proposes banning ‘superintelligence’ and putting violators in prison
The proposed US bill defines 'superintelligence' as capable of destroying or disempowering humanity and sets a 20-year prison term.
The pacing era's first launch day
Anthropic says Opus 5.5 matches Fable 5.1 at 40% lower cost, while OpenAI reports GPT-6 Sol and Luna halve prices of the models they replace.
Alibaba’s Qwen open-sources Qwen-Image-2.1 for unified image generation and editing
Qwen-Image-2.1's 7B visual component supports transparent generation, local edits, and composition with up to 10 reference images.
Shares of Chinese AI Model Firms Fall After Report of Regulatory Probe
Z.ai shares fell 12.4% after The Information reported China's internet regulator is probing potential data leaks to Anthropic.
ChatGPT Voice gets closer to "Her" with email, calendar, and Slack access
OpenAI says ChatGPT Voice now runs on GPT-6 Astra, Sol, and Luna and can access email, calendar, and Slack plugins.
Also covered byTechCrunch AIOpenAI extends cyber access to Ukraine for civilian defense
OpenAI states Ukraine's CERT-UA handled nearly 6,000 cyber incidents in 2025, motivating the Daybreak program extension.
Anthropic Says Its AI Helped Discover a Possible New Gene-Editing Tool
Anthropic claims Claude helped discover a previously unknown molecular system that could represent a new gene-editing tool.
Better prompt caching for GPT-6
OpenAI reports GPT-6 improves prompt caching with higher hit rates and new diagnostics to reduce latency and costs.
New York Times disputes extent of its errata in summary judgment statement but concedes 2 week extension to briefing may be warranted
The New York Times disputes OpenAI and Microsoft's characterization of its errata as extensive, calling some assertions false.
Use open weight models as your AI coding agent with Amazon Bedrock
AWS demonstrates using OpenCode with open weight models on Bedrock to keep proprietary code inside a user's own account.
Anthropic releases cheaper model in first launch since slowdown calls
Anthropic states Opus 5.5 matches or beats Fable 5.1 and Mythos 5.1 in coding, reasoning, and business workflows.
Meta’s Muse AI Assistant Rolled Out With a Serious Security Flaw
Meta says it fixed a Muse zero-day that gave attackers full control of a victim's Mac; Amazon began blocking Muse from its site.
Agentic conversational video intelligence built on AWS
AWS describes a single Strands Agents SDK agent orchestrating Bedrock, Rekognition, and Transcribe for video Q&A.
Pacing problem: can AI fears overcome US-China race dynamics and force a slowdown?
Anthropic, OpenAI, and SpaceXAI CEOs publicly call for pacing frontier AI development ahead of a US-China summit, while race dynamics persist.
Summary of METR's predeployment evaluation of Claude Opus 5.5
METR’s independent evaluation of Claude Opus 5.5 examines its impact on long‑horizon AI R&D tasks.
Claude discovers a novel enzyme system with CRISPR-like repeats
Anthropic's new life sciences lab used Claude to discover a novel enzyme system with CRISPR-like repeats.
Enabling Private High-Performance Production AI Inference with NVIDIA Confidential Computing
NVIDIA Confidential Computing enables secure production AI inference on encrypted hardware but introduces performance overhead.
Nokia Open-Sources AnyJev: A Training-Free Layer That Turns Any Open LLM Into a Calibrated Decision Model
Nokia open-sources AnyJev, a training-free library that turns any open LLM into a calibrated decision model by reading next-token probabilities.
Lightmatter launches an optical module that cuts AI cluster fiber in half
Lightmatter claims its single-strand optical module halves the fiber cabling needed for GPU clusters, but the company did not disclose specific deployment timelines.
MLCommons Joins EU-Funded AIRIS Project to Build and Benchmark Next-Generation Biomedical AI
MLCommons joins the EU-funded AIRIS project, which has €16.9 million to build mechanism-informed biomedical generative AI.
Muse is having a moment
Platformer reports Muse is gaining traction but questions whether consumer agents represent a durable shift or a mirage.
No stories match. Active filters.
Follow this briefing: Daily digest RSS.