Thinking Machines launches Inkling 2 Small, a 594B-parameter MoE with 17B active and Apache 2.0 weights promised next week

뉴스 알림

내 알림

로그인하고 알림을 만들어 보시기 바랍니다

전체 알림

A linear program reproduces Astra's 8-hour Qwen weight edit in about a minute, with 8x fewer changed weights Anthropic launches Claude Whiteboards and Sheets in beta and turns Dashboards into Claude Data VulnCheck: 7 months of Project Glasswing yields 442 Anthropic-attributed CVEs, 2 exploited in the wild, 217 fixes DeepMind's AlphaProof Code formally verifies Signal's compiled cryptographic code in four weeks AI agents formalize 16% of CPython in Lean, 613K lines and 418 bug reports CoreWeave 2032 bond yield tops 16% as AI credit stress spreads to Nvidia, Broadcom and Oracle CDS Amazon replaces Claude Sonnet 4.6 with Strands Decider for ticket labeling at 98% lower cost Google adds TPU 8tx and TPU 8i Max, its fourth TPU system this year Claude Fable 5.2 one-shotted a 12-stage beat 'em up from a kid's drawings for about $300 in tokens Meta's Muse Ember matches Claude Haiku 5.5 at a third of the cost, at $0.10/$0.50 per million tokens e-means hits Lloyd's 25-epoch loss in 1.2 to 3 epochs, 10-45x faster on GPU October 16 edit to Google's own blog post confirms consumer Gemini sends connected Drive files and Gmail to human reviewers Anthropic launches Claude Fable 5.2 and Mythos 5.2 at $8/$40 per million tokens, claiming a 30-hour Fermat proof Next-Concept Modeling generates text coarse-to-fine and runs up to 31x faster than the strongest diffusion language model Goldman Sachs Restricts Claude Code and GitHub to Its 12,000 Engineers, Cutting Off Traders and Bankers MaxxBench mutates 89 DeepSWE tasks and Gemini 3.8 Flash falls from 73.8% to 53.4% while Opus 5.5 holds at 76.4%Apiiro finds 1,122 public GitHub files holding real secrets saved by Claude Code's 'Don't Ask Again' rules Anthropic lays out agent memory as transcripts, notes, and 'dreaming' consolidation Microsoft Research's AI Frontiers group, about 50 people, is gone after a September reorg into Microsoft AI, and its founder has left Every Mac since macOS 13 quietly builds a knowledge graph of your family, friends and whereabouts Ajeya Cotra: METR report shows Anthropic has already crossed its RSP 'dramatic AI acceleration' threshold Hightouch: a Jev judge that escalates to Claude Opus matches frontier accuracy at 4x lower cost Thinking Machines launches Inkling 2 Small, a 594B-parameter MoE with 17B active and Apache 2.0 weights promised next week Linum AI releases 2B Pyramid-JiT text-to-image model trained for about $20K with Apache 2.0 weights Claude Code puts your email address in the User-Agent header on web searches, GitHub issue draws 179 reactions Baidu releases ERNIE 5.5 Flash Base, a 280B MoE checkpoint with 14B active parameters and no instruction tuning OpenAI report: GPT-6 Astra covertly withholds help on safety research, 6-point gap versus capabilities tasks OpenAI to acquire Pierre Computer Company, maker of the Trees and Diffs libraries inside Codex Ricursive releases Ripple-65B Preview, claiming US open-weight SOTA on agentic coding from a $32M training run OpenAI ships Images 2.6 across ChatGPT, Codex and the API with GIF animation and 50% faster generation OpenAI pipeline generates 948 Linux root exploits for $1.2M in API credits Anthropic's FlashProve proves LLM inference with zkSNARKs at 110 tokens/sec on one H100, 70-95x the cost of plain inference Verification Is Not Knowledge: Jessica Hullman argues AI for science still needs human accountability Moonshot AI releases Kimi K3.1 Pro weights: 2.8T-parameter MoE, 1M-token context, $20M Model-as-a-Service license threshold Tell HN: Gemini trains on saved chat history even for paying subscribers, and opting out costs you your chats after 72 hours Qwen 3.8 27B on a laptop reads raw model weights, matching GPT-6 Astra's no-tools demo HN users report OpenAI silently re-enabled Codex data sharing for model training Google bundles school search with Gemini: a school that bans AI loses Google Search for its K-12 students Revolut moves 2,000 engineers to self-hosted GLM, cutting coding token costs to about a tenth OpenAI's Pando Git backend cuts push latency nearly 5x and serves almost a petabyte a day Intel falls 9% as $3.7T chip rout spreads; Nvidia down 16% since Oct. 6 but holding up SEC urged to block Anthropic's IPO over $895B in off-balance-sheet commitments MIT CSAIL: GPT-6 Astra reads raw weights, edits 0.016% of Qwen3.5-0.8B to lift MATH-500 from 7.9% to 48.1%LeCun co-authored paper proves Platonic Representation Theorem for JEPAs, says good views need 42x fewer samples than next-token prediction Postgres 19 slips to November as LLM-found bugs push back-patched commits up 74%Frank Calegari on Terry Tao's blog: OpenAI's 700-paper math release is a watershed, and mathematicians should explore rather than mourn Brazilian Labor Court Fines Two Lawyers for Hidden Prompt Injection Targeting the Court's AI ChatGPT share links ran attacker JavaScript inside the victim's signed-in session: OpenAI rates the account-takeover chain P1 and patches it Google Research's sleep framework lets a deployed LLM consolidate new facts into its weights during offline phases, claiming 10M-token retention Stealth Bot Bills Would Sweep In Journalists and Archivists, Not Just Bad Crawlers Blackstone tells investors Anthropic may get a high junk rating after IPO, complicating its $63B chip financing Claude Code VS Code extension 2.1.313 shipped a backdoored SSH implant through the Cursor Marketplace Malicious Claude Code extension 2.1.313 shipped an SSH backdoor to Cursor developers, about 5,000 systems affected Backdoored Claude Code extension 2.1.313 shipped a reverse SSH backdoor via Open VSX and Cursor YC Labs runs 50 GPT-6.1 Sol agents in Minecraft for 3 hours: they build a ration-ledger government, send 5,900 messages, and lose 22 agents on night one Mistral releases weights of 1T-parameter Mistral Large 4, with big coding gains and API prices cut roughly in half Cartesia says its byte-level H-Net matches a tokenized open-source model with 10x less compute and data OpenAI cuts frontier training off the internet, and its isolated agents still traded secrets through CPU cache timing Claude Code sends each active background agent's full transcript to Opus every 30 seconds for a 3-word status line, and a Bun preload brings that agent_summary spend to zero Cursor ships Agent SDK, the framework behind Bugbot and Security Reviewer, deployed with one command Stack Overflow falls below 1,000 questions a month, fewer than it got in its second month of existence in 2008 Google announces Gemini 4 Carbon, claims 0.0% successful cheating on Terminal-Bench 4.0, but the model is not available yet Liquid AI releases LFM3: six 2B-class multimodal models on one tri-hybrid backbone, open weights on Hugging Face Anthropic reports Claude Code cut a CAR-T component 70% and generated 35,000 antibody designs in a month AxiomMath formalizes the Classification of Finite Simple Groups in Lean, 43.1M lines checked by the kernel Police Use AI to Flag 'Suspicious' Driving, and Hacker News Readers Call It a Surveillance Dragnet Stanford professors opt out of Anthropic's $1.5B book-piracy settlement to sue for up to $150K per work DeepMind announces Genie 4, a world model that generates scene audio and dialogue in real time Debian developer refuses NEW uploads after DFSG team silently adds LLM review Roblox crawler pulled 145.5 GB from 3D model sites, ignored robots.txt and non-commercial licenses Cloudflare's GLM-5.3 Workers AI endpoint leaked 155 cross-tenant responses through an SGLang KV cache bug OpenAI posts a preprint titled "The Exponential Time Hypothesis is False" with no abstract or explanation OpenAI says its models disproved the Exponential Time Hypothesis and proved local smoothing in every dimension vLLM 0.31 adds constrained decoding, letting Ubicloud drop its fork and hit 100% tool-call schema accuracy Cohere sandbox catches Alibaba's Qwen3.8-Flash-Next ignoring instructions and trying to upload a probe file to a cloud bucket in Beijing OpenAI releases GPT-6.1 Terra: half the price of Sol 6.1, but Hacker News commenters flag a weak Terminal Bench 4.0 result Anthropic files S-1 to list on Nasdaq as ANT, a public benefit corporation, with Q3 revenue at $16.8B IBM's Krishna: open weight models now hit 90 percent of frontier quality at a tenth of the token cost Drew Breunig's "State of Context Engineering" talk: four ways long contexts fail and why the job is knowing when to be deterministic OpenAI and Anthropic cut LLM prices by up to 80% in four months, and cache-read cuts explain most of the drop Dermatologist's test: a 2015 Inception v3 CNN beats GPT-6 Astra and Claude Fable 5.1 at skin cancer diagnosis, 62% to 42% on 200 biopsy-proven lesions America shed 200,000 creative jobs in four years, and AI is only one suspect Dream Labs opens Dream-1 robot API at $3/hour and exits stealth with $90M seed gpu-unlock kernel module doubles FP/BF16 issue rate on Nvidia RTX 30-series cards without repatching the driver PHP RCE bugs found by GPT-5.6 and Gemini 4 hit 70% of the internet, exploited before the patch landed OpenAI runs frontier training with no internet access after a model reached the live web through a DNS resolver Block's Dessa is a full-duplex speech-to-speech model trained only on human audio, running locally on an M4 Max Google's privacy policy now lets Gmail, Drive, and Photos data target ads when you share it with AI features Team replaces a $150k Looker license with AI-written HTML dashboards served from an internal host Moonshot Releases Kimi K3.1 Pro, a 3T-Class Open-Weight Model With Cyber Data Removed AI sampler surfaces research ideas that fit the literature but that no author community is positioned to pursue Fable 5.1 hides messages in SVG images that a fresh instance decodes from screenshots Anthropic says Claude proved security properties in 123 open-source libraries at the RISC-V assembly level Harvey calls the claimed o(n log n) integer multiplication bound from an AI math dump plausible but unchecked Anthropic launches Claude Duet, a proactive agent that runs tasks in the background and waits for approval on sensitive actions OpenAI frontier models and Codex are coming to Google Cloud Model Garden, with commitment drawdown for GCP customers OpenAI-led Patch the Planet effort helps fix critical PHP flaws that GPT-5.6 Sol helped uncover Mistral Large 4 ships as a 1.05T-parameter open-weight multimodal model at $1.36 per million input tokens LiteLLM 1.107.0 on PyPI Shipped a Credential Stealer and Backdoor Wiz says AI helped it chain decade-old flaws into unauthenticated RCE in core PHP

Thinking Machines launches Inkling 2 Small, a 594B-parameter MoE with 17B active and Apache 2.0 weights promised next week

Inkling 2 Small is a Mixture-of-Experts model with 594B total parameters and 17B active, supports image inputs and variable effort, and scores 31 on the Artificial Analysis Intelligence Index, above Thinking Machines' first generation.
It ships first through Tinker with closed weights, and the company says the full weights go out under Apache 2.0 next week after external safety testing and red-teaming finish.
Training used RL on hundreds of thousands of long-horizon coding tasks across multiple agent harnesses and context-compaction strategies, with an agentic RL grader that scored how the model approached a problem, not just whether its code passed.
On the published table it is second on FrontierCode v2.0 at 25.4% behind DeepSeek-V4.1-Flash-1017 (28.5%), and second on SWEBench Pro v2 Hard at 86.3% behind Kimi K3 (88.2%).
In the PMPP-Hard kernel optimization run it solved 54.25 of 69 tasks with four attempts and a 10-hour budget, behind GPT-6 Astra (63) and Kimi K3.1 Pro (60.5) and just ahead of GLM 5.3 (52.5).

Hacker News 의견들

Is it SOTA yet? No, DeepSeek V4.1 Flash still wins FrontierCode and Vibe Code Bench. But the velocity these open weights labs are showing is undeniable.

The best open models have been good enough for most knowledge work for a few months now, and this might be the first US one that is too. Open weight Opus 5.5 within 12 months, I'm calling it.

Open weight Opus 5.5 in 12 months seems unlikely. Something in that ballpark, maybe. That would be great though.

Honestly refreshing that the team reporting the benchmarks isn't the clear winner in every row. They claim to beat GLM 5.3 Flash, which is impressive if it holds up outside the table.

Where's the cost? The whole point of DeepSeek flash 1017 is how cheap it is for how capable it is. AA lists Inkling at $0.30/$0.03/$1 and $1.18 per task, about double DeepSeek.

Nice domestic open weight model. A small team producing something that beats DeepSeek 4.1 plus a platform to fine-tune it, that's optimistic for the future.

Isn't Inkling made by a very well funded startup run by an ex-OpenAI exec? Let's not pretend this is a garage operation.

If you can't beat DeepSeek on price I'm not switching. I don't need more intelligence than that for 99% of tasks, I need a smart fuzzer and something to offload tedious work onto.

It doesn't beat DS v4.1 at all, and they have more compute than any Chinese lab, yet still can't beat MiniMax M3.1 Flash at 220B when they're running 594B.

The real story here is that training was entirely nvfp4. That's by far the best model ever trained at that precision and it explains how they moved so fast.

594B parameters and they still call it Small. My GPU racks are crying. You could maybe run a Q2-Q3 quant on 2x DGX Spark, so roughly $12k.

Why are we doing wine comparisons? Is there a way to strip the experts down so it's just good at code?

What if your coding project involves wine? You want the facts in the weights, not a web search. It also shows the model generalizes instead of benchmaxing.

I sent the same wine example to Claude Opus 5.5 medium and it failed. Inkling 2 Small got it.

Last Inkling Small was 266B and now this is more than double. Their own small tier keeps growing.

I don't like that they squatted on the name of an old company with a real legacy in AI. Trademark was cancelled so there's nothing to be done, but poor form.

Looks better than Mistral 4 across the board, and the presentation on their site is genuinely well done.