https://gonioai.pages.dev/topics/models/Models & releases stories from gonioAI.enTue, 11 Aug 2026 10:45:13 +0000https://simonwillison.net/2026/Aug/10/introducing-muse-glimmer2026-08-11:models:https://simonwillison.net/2026/Aug/10/introducing-muse-glimmerTue, 11 Aug 2026 07:00:00 +0000Meta released Muse Glimmer, a dense 30B multimodal model under a clean Apache 2.0 license, logit-distilled from its larger Muse Spark and trained on agentic traces rather than the usual base-then-post-train recipe. It uses Gemma-4-style hybrid attention, quantizes to ~18GB at 4-bit (fitting a single 24GB GPU with a bundled DFlash speculative drafter), and ships a 128K native context that community testers stretched past 800K tokens with YaRN. Third-party benchmarks put it at 35 on Artificial Analysis's Intelligence Index, just behind Qwen3.6-27B; an open-weight Muse Spark 1.2 is promised within weeks. Zuckerberg paired the launch with a 6,000-word essay defending model distillation as 'learning from anything you can observe.' Why it matters: This is Meta's first open model since Llama 4 flopped, and a strong local-agent contender that directly needles OpenAI and Anthropic's anti-distillation lobbying. For self-hosters it fills the 24GB-GPU slot that Qwen3.6-27B and Gemma-4-31B couldn't.https://cactuscompute.com/needle2026-08-11:models:https://cactuscompute.com/needleTue, 11 Aug 2026 07:00:00 +0000Cactus released Needle 2, an Apache-2.0 45M-parameter model for tool calling, device control, and structured extraction that ships as a single 14MB binary running a full session in 28MB of RAM. Trained natively at 2-bit (CQ2) from pretraining onward rather than post-quantized, it hits 500 tok/s decode on a Raspberry Pi 5 and runs on ESP32-class microcontrollers. On five function-calling benchmarks (Mobile Actions, DroidCall, Seal-Tools, BFCL v4) it trades wins with LFM2.5-230M, FunctionGemma-270M, and Apple's Foundation Model at 5x to 70x smaller, though it lags on out-of-distribution Java/JavaScript and parallel calls. Pebble already runs it locally in its Index 01 ring app. Why it matters: It's a concrete bet that on-device tool-calling doesn't need billions of parameters or an NPU, aimed at the ~80% of edge devices that cost under $200. For anyone building always-on assistants, the confidence-score-driven escalate-to-cloud design is a clean private-by-default pattern.https://www.technologyreview.com/2026/08/10/1141511/these-startups-are-chasing-the-next-big-thing-in-llms2026-08-10:models:https://www.technologyreview.com/2026/08/10/1141511/these-startups-are-chasing-the-next-big-thing-in-llmsMon, 10 Aug 2026 07:00:00 +0000MIT Technology Review profiles a wave of startups attacking the transformer's dense-attention bottleneck. Subquadratic claims SubQ is the first sparse-attention mechanism to rival dense attention on search and coding; Manifest AI's 'power retention' keeps a rolling context summary, demoed via PowerCoder and Brumby; Liquid AI ships hybrid models that are 20% transformer, 80% liquid neural network and run on a Raspberry Pi; Inception's diffusion LLM Mercury 2 claims GPT-4-class quality at 10x speed; and Pathway's state-space Dragon Hatchling clears most of 250,000 hard sudoku that leading LLMs fail entirely. All the headline claims are self-reported and unverified, and industry skeptics remain. Why it matters: Dense attention is the main reason LLMs burn so much power and choke on long context. If any of these subquadratic approaches hold up outside a pitch deck, inference economics and context limits both move.https://the-decoder.com/google-dismantles-deepmind-and-bets-on-a-fresh-start-as-hassabis-heads-for-the-exit2026-08-09:models:https://the-decoder.com/google-dismantles-deepmind-and-bets-on-a-fresh-start-as-hassabis-heads-for-the-exitSun, 09 Aug 2026 07:00:00 +0000Following Jeff Dean's departure, reports say Google DeepMind is being downgraded to a subdivision: day-to-day operations pass to Koray Kavukcuoglu (without a CEO title), all Gemini work moves to the Bay Area, and Sergey Brin takes a larger role. Demis Hassabis was 'promoted' to chairman and could leave in the coming months to focus on Isomorphic Labs. SemiAnalysis reads the shakeup as Google conceding the frontier-model race and leaning into cloud and TPU revenue ($73B+ projected AI infra), while defenders frame it as a deliberate infrastructure play. Why it matters: The lab that produced the Transformer's successors and Gemini is being reorganized around cloud margins, not model leadership. If you build on Gemini, the roadmap signals matter: 3.1 Pro is still preview and 3.5 Pro appears shelved.https://the-decoder.com/googles-diffusiongemma-proves-you-dont-need-to-train-from-scratch-to-build-a-text-diffusion-model2026-08-09:models:https://the-decoder.com/googles-diffusiongemma-proves-you-dont-need-to-train-from-scratch-to-build-a-text-diffusion-modelSun, 09 Aug 2026 07:00:00 +0000Google DeepMind's technical report details how DiffusionGemma was built by converting Gemma-4-26B-A4B into a block-parallel diffusion model rather than training from scratch, using under 10% of the original token budget. It refines 256-token blocks in parallel at ~1,500 tokens/s on an H100, uses a combined RL-plus-sampler-distillation stage (SD·RL) that lifts reasoning benchmarks ~10 points, and can self-correct mid-derivation (near 85% on Sudoku after light tuning). Tradeoffs: it trails the autoregressive base in absolute quality, loops on repetition at aggressive step counts, and its speed edge collapses past ~32 concurrent requests. Apache 2.0 on Hugging Face. Why it matters: A recipe for turning existing open-weight autoregressive models into fast diffusion decoders is cheaper than training one, and the parallel self-correction is genuinely useful for structured outputs like JSON and code repair.https://www.reddit.com/r/LocalLLaMA/comments/1vjklwo/deepseek_v4_flash_0731_hits_827_on_terminalbench2026-08-09:models:https://www.reddit.com/r/LocalLLaMA/comments/1vjklwo/deepseek_v4_flash_0731_hits_827_on_terminalbenchSun, 09 Aug 2026 07:00:00 +0000DeepSeek reported 82.7% on Terminal-Bench 2.1 for V4 Flash 0731 using its unreleased 'DeepSeek Harness minimal mode.' The author of the Ante eval independently hit the same 82.7% (368/445 trials, ±1.79 SE) across 89 tasks at 5 trials each, max reasoning effort, no skills, via OpenRouter, with the full Harbor job public. The run confirms the model is highly harness-sensitive, echoing separate community results where switching agents (opencode vs pi) swung local-quant scores substantially. Why it matters: Independent reproduction of a vendor benchmark is rare and welcome, but the harness sensitivity is the real lesson: pick your agent framework carefully, because it can move scores more than the quant does.https://arstechnica.com/science/2026/08/deepminds-hurricane-model-bought-forecasters-an-extra-day2026-08-09:models:https://arstechnica.com/science/2026/08/deepminds-hurricane-model-bought-forecasters-an-extra-daySun, 09 Aug 2026 07:00:00 +0000A Nature paper shows Google DeepMind's WeatherNext model predicts cyclones with about a day more lead time than existing physics-based models, meaning its three-day forecasts match prior models' two-day accuracy. For 2025's Hurricane Melissa, it called a Category 5 Jamaica landfall with 80% confidence five days out, ahead of models that were still split on the track. Why it matters: One of the more concrete wins for ML weather models over numerical forecasting, on a task where an extra day of warning has direct human stakes rather than a benchmark number.https://openai.com/index/responding-next-frontier-critical-cyber-capabilities2026-08-08:models:https://openai.com/index/responding-next-frontier-critical-cyber-capabilitiesSat, 08 Aug 2026 07:00:00 +0000OpenAI says internal evals of its unreleased Astra model show such strong agentic-coding and cybersecurity gains that it 'cannot rule out' the Critical tier of its Preparedness Framework — the level where a model can find and chain zero-days against hardened targets with no human in the loop. It is pausing internal activities that lack safeguards and adding isolated test environments, weight encryption, and chain-of-thought monitoring; Sam Altman confirmed the rating will delay launch. Astra was not involved in the recent Hugging Face breach, and critics note OpenAI is flagging only the potential for a Critical rating, not the rating itself. Why it matters: First time a frontier lab has explicitly slowed a release over cyber risk — either a genuine capability inflection or well-timed 'too dangerous to ship' theater. Either way it sets the template for how labs gate agentic coding models.https://arstechnica.com/ai/2026/08/bytedance-trains-massive-ai-model-in-bid-to-rival-anthropic2026-08-08:models:https://arstechnica.com/ai/2026/08/bytedance-trains-massive-ai-model-in-bid-to-rival-anthropicSat, 08 Aug 2026 07:00:00 +0000Per the Financial Times, ByteDance is early in pre-training a model with as many as 10 trillion parameters — three times Moonshot's Kimi K3 and in the range of estimates for Anthropic's ~8T Mythos 5. Sources say ByteDance has avoided distillation from rival model outputs for over a year, and founder Zhang Yiming has told the 2,000-person Seed team to aim for world-leading capability. xAI is reportedly training 6T and 10T Grok variants on its Colossus 2 cluster. Why it matters: The parameter gap between Chinese labs and the US frontier is closing fast, and raw scale is back in fashion at the very moment everyone else is preaching the efficiency frontier.https://www.reddit.com/r/LocalLLaMA/comments/1vio0x6/deepseek_v4_flash_0731_appreciation_post2026-08-08:models:https://www.reddit.com/r/LocalLLaMA/comments/1vio0x6/deepseek_v4_flash_0731_appreciation_postSat, 08 Aug 2026 07:00:00 +0000DeepSeek's 304B MoE (6+1 active experts, native FP8, 1M context via sparse attention and KV compression) is drawing heavy local-deploy interest; Cline reported it became its most-used model with 3x token growth. Users on dual DGX Spark clock ~82 tok/s decode and praise it for hours-long coding and tool-use sessions, but a detailed writeup finds it loses nuance on summarization and speaker/pronoun tracking versus a much smaller Gemma-4-31B, and AMD MI325X users report broken tool-calling with the official vLLM recipe. Why it matters: A benchmark-topping open-weight MoE that shines on code and agents yet stumbles on office-text nuance — a reminder that intelligence-index scores don't predict what you actually deploy a model for.https://the-decoder.com/xais-imagine-image-2-0-lands-just-behind-openais-gpt-image-2-in-arena-benchmarks2026-08-08:models:https://the-decoder.com/xais-imagine-image-2-0-lands-just-behind-openais-gpt-image-2-in-arena-benchmarksSat, 08 Aug 2026 07:00:00 +0000xAI launched Imagine Image 2.0 as a 'Quality Mode' in Grok's web and mobile apps, adding a Magic Wand for localized edits, region segmentation, background removal, multi-reference editing (up to five inputs), and smart resize with generative fill. Its faster 'low' variant sits second on both Arena boards as of Aug 7 — 1,439 Elo in Image Edit and 1,320 in Text-to-Image — behind OpenAI's GPT-Image-2 (1,463 / 1,380) and ahead of Reve, Meta Muse-Image, Qwen-Image-3.0-Pro, Gemini and SeedDream. API access is 'coming soon.' Why it matters: The image-model leaderboard is now a genuine multi-way scrum; GPT-Image-2 still sets the bar, but no longer sits alone at the top.https://www.anthropic.com/news/improving-fable-5-s-biology-safeguards2026-08-07:models:https://www.anthropic.com/news/improving-fable-5-s-biology-safeguardsFri, 07 Aug 2026 07:00:00 +0000Anthropic rewrote the safety classifier's constitution for Claude Fable 5, cutting biology-related 'fallbacks'—where the system silently reroutes to the weaker Opus 5—by about 85% across product surfaces. Everyday health, lab-result, and educational queries should now stay on Fable 5, while dual-use areas like virology, toxicology, and molecular design still fall back. The company says total fallbacks drop roughly 67% on Claude.ai but only 17% in Claude Code and 7% on the API. Why it matters: If you build on Fable 5 and hit unexplained quality drops on benign science prompts, this is why—and the classifier margins mean false positives will persist, especially outside the consumer app.https://the-decoder.com/openai-improves-gpt-5-6-sol-in-chatgpt-and-restricts-free-users-to-its-weakest-model2026-08-07:models:https://the-decoder.com/openai-improves-gpt-5-6-sol-in-chatgpt-and-restricts-free-users-to-its-weakest-modelFri, 07 Aug 2026 07:00:00 +0000OpenAI merged 'Instant' and 'Thinking' into a single GPT-5.6 Sol for Plus/Pro users, adding a reasoning-effort slider, and claims 68% fewer factual-error responses than GPT-5.5 Instant on an internal finance/medicine/law eval. Free and Go users move to the smaller GPT-5.6 Luna with unlimited text chats and a 'Think' button—but no access to frontier reasoning. The changes apply only to ChatGPT; Sol in ChatGPT Work and Codex is unchanged. Why it matters: The unified model plus effort slider is the new default surface most users will hit, and the free-tier split makes 'ChatGPT said' an even less precise statement about which model actually answered.https://blogs.nvidia.com/blog/open-world-models-physical-ai2026-08-07:models:https://blogs.nvidia.com/blog/open-world-models-physical-aiFri, 07 Aug 2026 07:00:00 +0000NVIDIA released Cosmos 3, a mixture-of-transformers 'omni' family under the OpenMDW 1.1 license that combines vision reasoning, world generation, and action prediction in one stack. It comes in three sizes: Super (64B), Nano (16B), and Edge (4B) for on-device robot policy on Jetson and RTX GPUs. NVIDIA claims top open-weights rankings on Artificial Analysis for text-to-image and image-to-video, plus No. 1 on RoboLab for robot policy. Why it matters: World models that generate physically grounded synthetic data and simulate future states are the emerging substrate for robotics and AV teams, and open weights plus an Edge tier make specialization on your own hardware realistic.https://research.meta.ai/blog/introducing-muse-code-and-muse-spark-1-22026-08-06:models:https://research.meta.ai/blog/introducing-muse-code-and-muse-spark-1-2Thu, 06 Aug 2026 07:00:00 +0000Meta released Muse Code (beta), a terminal coding agent powered by the new Muse Spark 1.2 model, co-trained together so the model was tuned around the harness's toolset. Its runtime appends every model call, tool run and edit to a local event log for replay-exact, restart-safe recovery, and it fans big jobs out to persistent background sub-agents in isolated git worktrees. Muse Spark 1.2 is priced at $1.25/$4.25 per million input/output tokens, but a muse-spark-1.2-contributor tier drops to $0.10/$0.20 if you let Meta train on your data. Why it matters: Meta, long a coding-agent straggler, just matched Codex and Claude Code on architecture and undercut them on price — the resumable event log and persistent sub-agents are the parts other harness builders will copy.https://www.reddit.com/r/LocalLLaMA/comments/1vg569y/qwen_developers_responses_from_their_recent2026-08-06:models:https://www.reddit.com/r/LocalLLaMA/comments/1vg569y/qwen_developers_responses_from_their_recentThu, 06 Aug 2026 07:00:00 +0000In a developer AMA, the Qwen team confirmed the 2.4T-parameter, 95B-active Qwen3.8-Max (architecture similar to 3.5, scaled up) will get open weights, and that a brand-new Qwen3.8-27B — not a retrain of the 3.6 version — is coming with a 'pretty huge jump' in capability. A ModelScope listing points to a release next Wednesday. The team declined a technical report for this cycle, cited 'a truly unreasonable amount of compute' spent on post-training RL, and said Qwen now assists in nearly every stage of its own model iteration. Why it matters: A dense 27B that outperforms its predecessor plus open frontier-scale weights is exactly what local builders have been asking for, and the near-monthly cadence keeps pressure on both Chinese rivals and closed labs.https://huggingface.co/blog/LiquidAI/lfm2-5-2-6b2026-08-05:models:https://huggingface.co/blog/LiquidAI/lfm2-5-2-6bWed, 05 Aug 2026 07:00:00 +0000Liquid AI released LFM2.5-2.6B, a 2.69B-parameter model with 128K context and tool calling, post-trained specifically inside agent harnesses via SFT, teacher distillation, and agentic RL. The Q4_K_M GGUF is ~1.67GB and Liquid claims 30 tok/s on a phone, 113 tok/s on a Ryzen AI Max+ 395, and 220 tok/s on an M5 Max, in under 2.5GB. On tool-use benchmarks it edges Qwen3.5-9B (ToolSandbox 77.83 vs 76.44) but trails on coding (LiveCodeBench 59.41 vs 69.86); Liquid explicitly does not recommend it for agentic coding. Day-one support spans llama.cpp, MLX, vLLM, SGLang, and ONNX. Why it matters: The interesting use isn't a smarter assistant but cheap local worker agents doing extraction, search, and repetitive tool calls — though the 128K context and multi-turn stability claims still need independent testing.https://qwen.ai/blog?id=qwen3.82026-08-03:models:https://qwen.ai/blog?id=qwen3.8Mon, 03 Aug 2026 07:00:00 +0000Alibaba released Qwen3.8-Max, its largest model yet at 2.4 trillion parameters, sharing benchmark results that rank it above Moonshot's Kimi K3 and comparable to or better than Anthropic's Fable 5 on several tests. A smaller Qwen3.8-27B was announced alongside it; Unsloth's Daniel Han says the 27B fits in about 17GB of VRAM. The Max numbers are Alibaba's own, so treat the Fable 5 comparison as a vendor claim until third parties replicate it. Why it matters: Another Chinese lab is claiming frontier-parity within weeks of Kimi K3, and the paired 27B means the same generation is usable on a single consumer GPU, not just via API.https://www.interconnects.ai/p/latest-open-artifacts-23-laguna-s212026-08-03:models:https://www.interconnects.ai/p/latest-open-artifacts-23-laguna-s21Mon, 03 Aug 2026 07:00:00 +0000Poolside pushed a fully re-trained Laguna-S-2.1 checkpoint (118B-A8B, fits on a DGX Spark) under the OpenMDW license, its third Artifacts appearance in three months. Interconnects' latest open-models recap frames the moment as sustained proliferation rather than the long-predicted consolidation, spanning Thinking Machines' Inkling, Tencent's Apache-2.0 Hy3, Meituan's 1.6T LongCat-2.0 trained entirely on Ascend 910s, and DeepSeek-V4-Flash-0731 edging Laguna on the frontier. Note the licensing catch: Kimi K3-style revenue-share terms may expose US firms to future policy action. Why it matters: The bet has flipped from 'labs will consolidate' to 'more labs keep shipping open weights' — good for builders, but the licenses are getting geopolitically loaded.https://the-decoder.com/openai-announces-its-next-major-model-astra-by-dropping-ten-previously-unsolved-math-solutions2026-08-01:models:https://the-decoder.com/openai-announces-its-next-major-model-astra-by-dropping-ten-previously-unsolved-math-solutionsSat, 01 Aug 2026 07:00:00 +0000OpenAI previewed Astra, a next-gen model family built to coordinate multiple agents over hours or days, and published a report claiming an internal version solved ten previously open problems in math and theoretical CS, spanning group theory (the existence of non-sofic groups), lattice cryptography, coding theory and quantum complexity. Each proof was formalized in Lean for machine-checking, and OpenAI says the tokens cost roughly $2,000 per solution at Sol API rates. Astra is slated to be the first model submitted to the Trump administration's planned pre-release federal review. Why it matters: The Lean-formalized proofs are a concrete, verifiable capability claim rather than a benchmark number, but mathematicians note Astra was trained on essentially all of human mathematics and cracked no Millennium Prize problems, so calibrate the hype accordingly.https://mashable.com/tech/anthropic-releases-claude-opus-52026-08-01:models:https://mashable.com/tech/anthropic-releases-claude-opus-5Sat, 01 Aug 2026 07:00:00 +0000Anthropic released Claude Opus 5 at $5/$25 per million input/output tokens (same as Opus 4.8) and made it the default on Claude Max. It claims intelligence close to Fable 5 at half the price, the lowest deceptiveness rates of any Anthropic model, and wins over GPT-5.6 Sol on every benchmark except agentic coding. Notably, Anthropic says it deliberately left offensive-cyber tasks out of training, so Opus 5 can find vulnerabilities but is much worse at exploiting them than Mythos and older models. Why it matters: The intentional cyber nerf is a pointed design choice given the week's containment incidents, and a rare case of a lab shipping a model that is deliberately less capable at something.https://the-decoder.com/thinking-machines-bets-on-efficiency-over-size-with-its-second-model-inkling-small2026-08-01:models:https://the-decoder.com/thinking-machines-bets-on-efficiency-over-size-with-its-second-model-inkling-smallSat, 01 Aug 2026 07:00:00 +0000Mira Murati's Thinking Machines released Inkling Small, an Apache 2.0 open-weights reasoning model with 276B total and 12B active parameters. Artificial Analysis scores it 40 on the Intelligence Index, one point below the larger Inkling, and says no open model of equal or smaller size scores higher. It beats its bigger sibling on some coding and reasoning tests while averaging 24K output tokens per task, versus 45K for DeepSeek V4 Flash and 78K for GPT-5.4 mini. It handles text, image and speech, has a 256K context window, and is fine-tunable in-browser via Tinker Playground. Why it matters: The token-efficiency gap is the real story: at a third of Inkling's parameters and roughly half the output tokens of rivals, Inkling Small is a cheaper base to fine-tune on your own data.https://www.latent.space/p/ainews-gpt-56-price-cut-by-20-802026-07-31:models:https://www.latent.space/p/ainews-gpt-56-price-cut-by-20-80Fri, 31 Jul 2026 07:00:00 +0000OpenAI dropped GPT-5.6 Luna 80% (now $0.20/$1.20 per million in/out tokens) and Terra 20% ($2/$12), and added a Sol Fast tier running up to 2.5x lower latency at 2x price with no claimed intelligence change. The company attributes the cuts to systems work partly done by GPT-5.6 Sol itself, which it says analyzed production traffic and autonomously rewrote Triton and Gluon serving kernels to cut end-to-end costs ~20%, plus a >15% speculative-decoding gain. Swyx's analysis notes GPT-5.4's full flagship intelligence (AA index 51) now sells at roughly one-thirteenth of March's token price via Luna, and OpenAI is moving Codex and ChatGPT auto-review off GPT-5.4 onto Luna for ~10x lower cost. Why it matters: Constant-level intelligence is getting an order of magnitude cheaper every few months, and OpenAI now undercuts several open models on cost-per-task. For anyone budgeting agent workloads, re-pricing your stack quarterly is no longer optional.https://www.reddit.com/r/LocalLLaMA/comments/1vbdsmz/minimaxh3_video_model_released_open_weights2026-07-31:models:https://www.reddit.com/r/LocalLLaMA/comments/1vbdsmz/minimaxh3_video_model_released_open_weightsFri, 31 Jul 2026 07:00:00 +0000MiniMax launched H3, a multimodal model that generates up to 15 seconds of 2K video with native stereo audio, plus video-to-video motion transfer and text/brand rendering aimed at commercial content. On Artificial Analysis it leads video editing and beats ByteDance's Seedance 2.0 in some tasks, but trails Google's Gemini Omni Flash on text-to-video and sits behind both on image-to-video. MiniMax says 2K pricing is under a third of mainstream models' rates and plans to release the weights 'in the coming days' under the MiniMax Community License, which permits free non-commercial use and commercial use for organizations under $20M revenue with attribution. Why it matters: Open weights have barely touched video generation, which remains closed-source and slow-iterating. If H3's weights actually ship at these prices, it's the first credible open base for teams building video pipelines instead of renting an API.https://www.reddit.com/r/LocalLLaMA/comments/1vbimop/deepseek_v4_flash_has_a_nice_bump_in_capability2026-07-31:models:https://www.reddit.com/r/LocalLLaMA/comments/1vbimop/deepseek_v4_flash_has_a_nice_bump_in_capabilityFri, 31 Jul 2026 07:00:00 +0000DeepSeek's V4 Flash is now live on the API, with V4 Pro promised 'soon.' The 0731 release posts sharp gains over the earlier preview: Terminal Bench 56.9 to 82.7 (on a shifted v2.0-to-v2.1 suite) and Toolathlon 51.8 to 70.3, plus new scores on NL2Repo, DeepSWE and Cybergym. Against GPT-5.6 Terra it trades blows, leading Toolathlon by 17 points but trailing on DeepSWE and Agents' Last Exam. On the Artificial Analysis Intelligence Index it lands at 50, one point behind GLM-5.2 and GPT-5.6 Luna. Why it matters: Flash is DeepSeek's cheap tier, and it's now within a point of frontier-adjacent models on the aggregate index while leading on some tool-use benchmarks. It sharpens the pressure OpenAI's price cuts were reacting to.https://www.reddit.com/r/LocalLLaMA/comments/1vbj6uf/huawei_opensouced_openpangu20pro_505ba18b2026-07-31:models:https://www.reddit.com/r/LocalLLaMA/comments/1vbj6uf/huawei_opensouced_openpangu20pro_505ba18bFri, 31 Jul 2026 07:00:00 +0000Huawei open-sourced openPangu-2.0-Pro, a 505B-parameter MoE (18B active) with 512k context, pretrained on 34T tokens and trained entirely on Ascend hardware. LG AI Research released K-EXAONE 2.0 under Apache 2.0, a 750B-A37B model (3x its 236B v1) covering 10 languages and built under Korea's Sovereign AI project, reporting long-context and agentic tool-use scores ahead of Qwen 3.5 and GLM-5.1 on their own benchmarks. Both land as a permissively licensed alternative to the frontier API tier. Why it matters: The open-weights cadence out of Asia isn't slowing, and Ascend-trained and Apache-licensed drops matter for teams that need sovereignty or want off the NVIDIA-and-OpenAI treadmill. As always, treat the self-reported benchmarks with suspicion until independent runs land.https://techcrunch.com/2026/07/27/microsoft-launches-its-first-cyber-model-and-a-new-agentic-cybersecurity-system2026-07-28:models:https://techcrunch.com/2026/07/27/microsoft-launches-its-first-cyber-model-and-a-new-agentic-cybersecurity-systemTue, 28 Jul 2026 07:00:00 +0000Microsoft launched MAI-Cyber-1-Flash, a compact security model derived from its MAI-Thinking-1 line, wired into its MDASH multi-agent vulnerability harness. The combined system scores 96% on CyberGym (+12 points over Anthropic's Mythos, and ahead of Gemini and GPT), with Microsoft claiming a 50% cost cut by having the Flash model handle ~90% of tasks and escalating the toughest 10% to GPT-5.4. It also unveiled Perception, an agentic platform of red/blue/green teams, in preview November 3. Why it matters: Microsoft is positioning itself as a model orchestrator rather than a single-model shop, and the cheap-worker-plus-frontier-escalation pattern is becoming the default architecture for cost-sensitive agentic workloads.https://simonwillison.net/2026/Jul/27/kimi-k32026-07-28:models:https://simonwillison.net/2026/Jul/27/kimi-k3Tue, 28 Jul 2026 07:00:00 +0000Now that Moonshot's 2.8T-parameter K3 is actually on Hugging Face (1.56TB, MXFP4), the details matter. The license isn't MIT/Apache: any Model-as-a-Service business over $20M revenue in a rolling 12 months must sign a separate agreement, and Moonshot pointedly calls it 'open weight,' not open source. Deployment math is brutal—104B active params won't fit on a 512GB Mac Studio, and even 8xH200 needs two nodes; only 8xB300 fits it single-node with KV cache. OpenRouter already lists K3 from seven providers, mostly at Moonshot's own $3/$15 per million tokens. Why it matters: The best open-weight model in the world ships with commercial carve-outs and server-class hardware requirements, a useful signal for where 'open' frontier models are actually settling: source-available, not OSI-licensed, and not something you run at home.https://www.reddit.com/r/LocalLLaMA/comments/1v722bp/kimi_k3_gets_open_weighted_tomorrow2026-07-27:models:https://www.reddit.com/r/LocalLLaMA/comments/1v722bp/kimi_k3_gets_open_weighted_tomorrowMon, 27 Jul 2026 07:00:00 +0000Moonshot AI is releasing the weights for Kimi K3, a 2.8-trillion-parameter mixture-of-experts model that launched as an API on July 16 and drew praise for coding, reasoning and agentic work. Founder Yang Zhilin is pitching openness and availability as the wedge against proprietary US systems. The catch for this crowd: at 2.8T parameters almost nobody can self-host it, so the practical near-term win is third-party inference providers rather than local runs. Why it matters: A genuinely frontier-class model going open-weight resets the price floor and hands distillation and fine-tuning targets to everyone; the hard part is now inference economics, not access.https://the-decoder.com/anthropics-opus-5-blows-past-fable-5-and-gpt-5-6-sol-on-the-benchmark-designed-to-measure-real-intelligence2026-07-26:models:https://the-decoder.com/anthropics-opus-5-blows-past-fable-5-and-gpt-5-6-sol-on-the-benchmark-designed-to-measure-real-intelligenceSun, 26 Jul 2026 07:00:00 +0000Claude Opus 5 scored 30.2 percent on ARC-AGI-3, up from the prior record of 7.8 percent set by GPT-5.6 Sol (Max), and solved five previously unsolved environments. ARC Prize credits genuine reasoning gains: the model translated tasks into algebraic notation and derived reflection equations unprompted. On the saturated older tests it merely matches the field (90.4 percent on ARC-AGI-2, 97.5 percent on ARC-AGI-1, at higher cost). Separately, Anthropic reports a 0 percent prompt-injection success rate across 129 browser-agent scenarios, but only with Cowork's two Auto Mode defense layers on; the bare model sits at 3.7 percent. Why it matters: Benchmark leaps this large usually mean targeted training. The tell: Opus 5 was built after ARC-AGI-3 went public, and a private test (Witness) shows much narrower gains.