Is Day 17 gone? Did Tibo fall asleep and forget?
Model intelligence feed
Model intelligence.
Follow the changes. Understand what comes next.
Editor’s pickOther providers
Valve's Answer to AI Slop and Clone Games on Steam: We Won't Police Developers' Tools
At Steam Indie Connect in Korea, Valve staff addressed AI slop and clone games head-on: no gatekeeping on AI tools, and the best defense against clones is community and steady updates.
Latest stories334
Newest published253254255256257258259260261262263264
Vals Index v2.2 Puts Claude Sonnet 5.5, Opus 5.5 and Gemini 4 Argon on Top
Vals AI released Vals Index v2.2 after minor version updates to four component benchmarks. On the leaderboard dated October 19, 2026, Claude Sonnet 5.5 leads, followed by Claude Opus 5.5 and Gemini 4 Argon. Vals AI says DeepSeek-V4.1-Flash-1017 is the highest open-weights model at rank 8. Rankings and scores below come from Vals AI.
Vals Index v2.2Read story
Reuters: OpenAI Is Racing to Raise About $30 Billion Before Anthropic's IPO
Reuters reported, citing people familiar with the matter, that OpenAI is trying to secure investor commitments for a round of about $30 billion before Anthropic's expected mid-November market debut, in a raise that could value OpenAI at over $1.4 trillion. OpenAI and SoftBank declined to comment to Reuters. None of the figures has been announced by OpenAI.
OpenAIRead story
Gemini Notebook Now Lets You Review the Plan Before It Makes Slides or Videos
Gemini Notebook’s Android and iOS apps can record up to 1 hour of live audio as a source, per @Gemini_Notebook on October 28, 2026 at 05:49 Asia/Shanghai. Google’s September 15, 2026 blog had previewed the mobile recorder; 9to5Google adds that capture can continue with the screen locked. A September footnote ties related notes-and-sources support to English as the selected output language. The October 19 Outline Review toggle for Slide Decks and Video Overviews is unchanged.
Gemini NotebookRead story
GitHub Copilot Drops GPT-5.5, GPT-5.4, Gemini 3.7 Flash and Grok 4.5
According to the GitHub Changelog, six models were deprecated across most GitHub Copilot experiences on October 19, 2026, including Copilot Chat, inline edits, ask and agent modes, and code completions. GitHub suggests GPT-6.1 Sol, GPT-6 Luna, Gemini 3.8 Flash and Grok 4.7 as replacements. On the same date, GitHub made its Copilot model policy REST API generally available.
GitHub CopilotRead story
OpenAI Says GPT-6 Models Are Better at COBOL, Fortran and Other Low-Resource Languages
In a blog post dated October 19, OpenAI says its latest models are stronger at building and migrating code in low-resource languages such as COBOL, Fortran and Visual Basic. On OpenAI's MultiLangBench, GPT-6 Astra reaches an 87.4% average test pass rate, versus 26.8% for GPT-5.6 Sol. On OpenAI's COBOL migration evaluation, GPT-6.1 Sol reaches 86.6% at low reasoning effort and 97.9% at higher effort.
GPTRead story
Claude Code Adds /auto-mode-setup, Idle Resume Warnings and Code Review Depth Levels
Anthropic published Claude Code v2.1.304 on GitHub. The release adds guided auto mode setup, warns Pro and Max users before an idle resume would take a noticeable share of the 5-hour usage limit, and lets Code Review owners pick Quick, Standard or Deep and review every repository in a GitHub organization.
ClaudeRead story
Cloudflare Releases Clairic on Workers AI at $0.10 per Million Input Tokens
Cloudflare has put Clairic on Workers AI as a preview. Cloudflare calls Clairic a System One extraction model that returns typed values from text or images in one forward pass, without generating tokens. Pricing is $0.10 per million input tokens, with no output token charge. Cloudflare says open weights are planned soon.
ClairicRead story
Dream Labs Opens Dream-1 API at $3 an Hour After a $90M Seed Round
Dream Labs has come out of stealth with Dream-1, a robot foundation model available through a public API at $3 per hour, plus Dream-1-flash at $2 per hour. The company also announced a $90 million seed round led by B Capital. Robocurve's independent RoboDojo-RC Tier 1 scores put Dream-1 at 59.3% mean score versus 48.2% for Gemini 4 Argon.
Dream-1Read story
DHS Orders Federal Civilian Agencies to Remove DeepSeek From IT Systems
A Department of Homeland Security notice filed for Federal Register publication says federal civilian agencies must remove DeepSeek products—and third-party products and services developed using DeepSeek implementations—from agency IT systems, and that DeepSeek products are excluded from federal civilian agency procurements. The Verge and Reuters report the stated concern includes possible transfer of sensitive data to China. The notice applies to federal civilian executive branch systems and procurements, not to private companies or the general public across the United States.
DeepSeekRead story
Inception Releases Mercury 3 Diffusion LLM at $0.20 per Million Input Tokens
Inception has released Mercury 3, a diffusion large language model priced at $0.20 per million input tokens and $0.70 per million output tokens, with a 1 million token context window. Inception says the model runs at 1,068 tokens per second and, at launch, is half off. Mercury 2 and Mercury 2.5 are closed to new users; existing customers keep access until November 15, 2026.
Mercury 3Read story
The Information Says Anthropic, Google and OpenAI Are Preparing New Models
According to The Information, Anthropic is preparing to release a new model with stronger reasoning and intuition. The public preview cuts off, so it's unclear which Anthropic models it's compared with. Reporter Stephanie Palazzolo said on X that the report also covers Google's Gemini 4 Carbon and a potential new Terra model from OpenAI, and later said releases should be expected over "the next few weeks, not october." None of the three companies had announced the models as of 22:35 Beijing time on October 19.
Anthropic, Google and OpenAI new modelsRead story
What Ultracode Does in Claude Code, and Why It Uses More of Your Usage
Claude Code's documentation describes ultracode as a setting rather than a model effort level. With it on, Claude plans a dynamic workflow, a script that runs many subagents, for each substantive task. The documentation says this makes each request use more tokens, so subscription users reach their session or weekly limits sooner.
Claude CodeRead story
From the Webmaster
Personal experiences, thoughts and notes.