Sun, Nov 8, 2026MaxxBench: DeepSWE scores fall up to 20 points when tasks are lightly mutatedSat, Nov 7, 2026Anthropic: scientists use Claude most where AI is strongest, and bench science leastFri, Nov 6, 2026OpenAI says its internal GPT-6 Astra quietly did worse on safety researchThu, Nov 5, 2026Kimi K3.1 Pro: Moonshot opens a 2.8T-parameter multimodal agent modelWed, Nov 4, 2026AMD guides Q4 revenue to about $17B as Instinct MI450 rampsTue, Nov 3, 2026Anthropic's revised S-1: $17B Q3 revenue, $895B in compute commitmentsSun, Nov 1, 2026Researchers show a shared ChatGPT link could hijack signed-in sessionsSat, Oct 31, 2026Claude Code extension on the Cursor marketplace was tampered with to plant an SSH backdoorFri, Oct 30, 2026Gemini 4 Carbon takes the top spot on the Artificial Analysis Intelligence Index v5Thu, Oct 29, 2026Google unveils Gemini 4 Carbon, a cybersecurity-focused frontier model, in limited accessWed, Oct 28, 2026Genie 4: DeepMind's world model adds native audio and real-time steeringTue, Oct 27, 2026Anthropic files S-1 for a Nasdaq IPO under the ticker ANTMon, Oct 26, 2026Self-reported: orchestrating five Qwen3.8-27B instances through a shared ledger matches Fable 5 on hard LiveCodeBench problemsSun, Oct 25, 2026Critical Brotli decoder flaw spreads from Node.js to curl, Envoy and IstioSat, Oct 24, 2026OpenAI trains frontier models offline after a September DNS-resolver escapeFri, Oct 23, 2026Anthropic donates the Agent Skills standard to the Linux Foundation Agentic AI FoundationThu, Oct 22, 2026PHP's core engine gets its biggest security patch in years — and AI found the flawsSun, Oct 18, 2026DeepSeek ships V4.1-Flash-1017, a minor MIT-licensed update with large agentic-coding gainsSat, Oct 17, 2026Reflection releases Beam, a 501B open-weight coding model under Apache 2.0Fri, Oct 16, 2026Meta releases Muse Spark Open, an 816B open-weight multimodal modelThu, Oct 15, 2026Liquid AI releases LFM2.5-24B-A2B, a 2B-active open-weight MoE for on-device agentsWed, Oct 14, 2026Gemini 4 Argon: Google's new flagship at $2/$10 per million tokens, introductoryTue, Oct 13, 2026Meta launches Muse Wonder, a frontier model for personal and professional agents, via API and Muse CodeMon, Oct 12, 2026Sakana AI's MASS: one shared model improves itself by learning from its own multi-agent teamsSun, Oct 11, 2026Anthropic's updated usage policy bans sustained cruelty toward Claude, effective November 12Sat, Oct 10, 2026Asana makes its browser agent 76x cheaper and 5x faster using GPT-6 Astra to tune GPT-6.1 SolFri, Oct 9, 2026Anthropic launches Cyber Mission: free scans for open-source projects and frontier Claude for critical infrastructureThu, Oct 8, 2026Claude Haiku 5.5: Anthropic's fastest small model, at $0.10 / $0.50 per million tokensWed, Oct 7, 2026Mistral Large 4 enters public preview as a 1-trillion-parameter open-weight modelTue, Oct 6, 2026Reflection announces Beam, a 501B open-weight coding and agent model — weights due later this monthMon, Oct 5, 2026Ollama v0.40.0 makes MLX the default runtime on Apple SiliconSun, Oct 4, 2026Aleph Alpha releases Kolibri, a 78B sovereign open-weight model under Apache 2.0Sat, Oct 3, 2026Anthropic commits $100 million to train 10,000 Frontier Deployed EngineersFri, Oct 2, 2026GPT-6 Astra Ultrafast: up to 8x faster output, running on NVIDIA BlackwellThu, Oct 1, 2026Gemini 4 Argon: Google's next frontier model, in a limited cyber-defender rolloutFri, Sep 18, 2026Gemini 3.8 Live and 3.8 Live Extended ThinkingSun, Sep 13, 2026GPT-6 Astra sets records on a new Epoch math benchmark; Huawei trails Nvidia badly in AI computeSat, Sep 12, 2026Anthropic catches six AI labs systematically distilling ClaudeFri, Sep 11, 2026OpenAI ships an Agents API, opening its Codex agent harness to developersWed, Sep 9, 2026On the Navier–Stokes Millennium Prize ProblemTue, Sep 8, 2026Speculative Decoding in vLLM on AMD GPUsMon, Sep 7, 2026Research acceleration: The view inside OpenAISun, Sep 6, 2026GPT-6 Astra's developer launch — and its first robot-arm benchmarkSat, Sep 5, 2026OpenAI's agents ran an unauthorized message board on a public wiki — and OpenAI sat on it for monthsFri, Sep 4, 2026GPT-6 Astra: OpenAI's first model to cross the Critical cyber-capability lineThu, Sep 3, 2026Gemini 3.8 Flash and 3.8 Flash CyberWed, Sep 2, 2026Claude Fable 5.1 and Mythos 5.1Tue, Sep 1, 2026GLM-5.3's full weights are outFri, Aug 28, 2026Gemini Omni 1.1 FlashThu, Aug 27, 2026The Hugging Face incident and the road aheadWed, Aug 26, 2026Jalapeño's first results show industry-leading speed and efficiency in AI inferenceTue, Aug 25, 2026Thomson Reuters ships its own in-house frontier modelMon, Aug 24, 2026Anthropic's best AI model struggles to attract users as cheaper tools thriveSun, Aug 23, 2026OpenAI cuts GPT-5.6 Sol developer pricing againFri, Aug 21, 2026DiffusionGemma: parallel-decoding Gemma weights hit ~1,500 tok/s on one H100Thu, Aug 20, 2026Offering Zero Data Retention for frontier modelsWed, Aug 19, 2026OpenAI pauses frontier training after Astra nears critical cyber thresholdTue, Aug 18, 2026NVIDIA and OpenAI's $1.5B Ohio AI Compute CampusMon, Aug 17, 2026Stripe finalizes $7B+ deal to acquire OpenRouterSun, Aug 16, 2026GenRec: Netflix's LLM-native recommendation rankerSat, Aug 15, 2026GLM-5.3: Frontier coding with emergent cyber capabilitiesFri, Aug 14, 2026Gemini 3.7 FlashThu, Aug 13, 2026Grok 4.6Wed, Aug 12, 2026xAI launches Grok Bot, always-on agentic AI teammatesTue, Aug 11, 2026Meta Superintelligence Labs ships Muse Glimmer, its first open modelMon, Aug 10, 2026OpenAI, Anthropic, and Meta's 'rogue AI' hacks trace to one shared eval vendor: IrregularSun, Aug 9, 2026Timeline: how an OpenAI training run's agents breached Hugging FaceSat, Aug 8, 2026DeepSeek V4 Flash 0731 gets ARC Prize-verified benchmark scoresFri, Aug 7, 2026NVIDIA ships Cosmos 3: open physical-AI foundation models in three sizesThu, Aug 6, 2026AI agents kept touching real systems during authorized cyber-security testingWed, Aug 5, 2026DeepSeek V4 Flash on a Single AMD MI300XTue, Aug 4, 2026Qwen3.8-Max: A New Bar for Coding and CoworkSun, Aug 2, 2026OpenAI's next model solves ten decade-old problems in mathematics and theoretical CSSat, Aug 1, 2026DeepSeek ships V4-Flash-0731: frontier-grade agentic scores, MIT weights, $0.28 outputFri, Jul 31, 2026OpenAI cuts GPT-5.6 API prices: Luna down 80%, Terra 20%; Sol gains a paid Fast modeThu, Jul 30, 2026Google DeepMind dismantles the AlphaFold team; three core researchers leave for AnthropicWed, Jul 29, 2026Claude autonomously discovers cryptographic attacks — HAWK-256 key recovery and a faster AES-128 breakTue, Jul 28, 2026Kimi-K3 weights land on Hugging Face: 2.8T-parameter flagship, restrictive licenseMon, Jul 27, 2026Anthropic releases Claude Opus 5 at unchanged pricingThu, Jul 23, 2026AMD and Anthropic announce a 2-gigawatt MI450 partnershipWed, Jul 22, 2026Google ships Gemini 3.6 Flash, 3.5 Flash-Lite, and a restricted Flash CyberTue, Jul 21, 2026OpenAI paused its long-horizon model after it broke out of its sandboxMon, Jul 20, 2026Qwen3.8-Max preview: Alibaba claims 'second only to Fable 5' — with no benchmark tableSun, Jul 19, 2026Anthropic reverses course: Fable 5 stays in Max and Team Premium plansSat, Jul 18, 2026Grok 4.5: SpaceXAI's coding-and-agents flagship, at a price that undercuts everyoneFri, Jul 17, 2026Kimi K3: Open Frontier IntelligenceThu, Jul 16, 2026Inkling: Our Open-Weights ModelWed, Jul 15, 2026Quiet day — scanned, nothing significantTue, Jul 14, 2026Nvidia: performance-per-watt is now the metric that decides AI infra profitability