MODELS & RELEASES
Every Models & Releases story from RXed AI News over the last 30 days — 50 items.
Meta upgrades AI chatbot with productivity features, including calendar access, to compete with Gemini, ChatGPT, and Claude. Source: theverge.com/tech/970570/meta-ai-chatbot-productivi…
Anthropic releases Opus 5, claiming capabilities 'close' to Fable 5. Launch follows US government clash and OpenAI security incident. Source: theverge.com/ai-artificial-intelligence/970105/clau…
@PoolsideAI releases Laguna S 2.1, an open-weight agentic coding model matching larger models on SWE-Bench Multilingual.
marktechpost.com/2026/07/21/poolside-releases-lagun…
Google launches Gemini 3.5 Flash Cyber, a cost-efficient AI security model for rapid vulnerability detection and patching. Source: theverge.com/tech/968572/google-gemini-flash-cyber-…
Moonshot's Kimi K3 leads Code Arena: Frontend but trails in advanced math. the-decoder.com/moonshots-kimi-k3-outperforms-fable…
Alibaba's Qwen 3.8 is a 2.4T-parameter multimodal model the team says trails only Fable 5. the-decoder.com/alibabas-qwen-takes-on-kimi-k3-with…
Developer fine-tunes @OpenBMB's MiniCPM5-1B on Claude traces to ship a 657MB local thinking model.
marktechpost.com/2026/07/19/someone-fine-tuned-open…
Open-weight models like @GLM4 and @DeepSeek_V4 now match @OpenAI's cyber capabilities from four months ago at a fraction of the cost.
the-decoder.com/open-weight-models-now-match-fronti…
Palm Garden AI develops Coherence Guard relational decision layer for human-facing robots.
therobotreport.com/palm-garden-ai-develops-coherenc…
@Zyphra released ZUNA1.1, an Apache 2.0 EEG foundation model with variable-length inputs from 0.5 to 30 seconds.
marktechpost.com/2026/07/17/zyphra-releases-zuna1-1…
A 27B model just hit 90% of full performance under 4GB — on a phone. The same week, two labs shipped trillion-parameter open models. Which side wins depends on what you're building.
rxed.ai/blog/2026-07-18-pocket-sized-frontier.html
Free Sunday briefing: subscribe.rxed.ai
China's Kimi K3 matches Anthropic's Opus 4.8 with a 300-person team, forcing Western labs to question their compute advantage. @the-decoder.com/just-like-deepseek-chinas-kimi-k3-is-forcing-western-ai-labs-to-question-their-compute-advantage/
@KimiAI's open-weight model K3 approaches GPT-5.6 Sol and Claude Fable 5 with 2.8 trillion parameters.
the-decoder.com/kimis-open-model-k3-nears-gpt-5-6-s…
Soofi Consortium releases Soofi S 30B-A3B, an open hybrid Mamba-Transformer MoE model for German and English.
marktechpost.com/2026/07/15/soofi-consortium-releas…
Moonshot’s Kimi 3, set to launch with 2–3 trillion parameters, aims to narrow the gap with Anthropic’s Opus 4.8, per FT. Source: techcrunch.com/2026/07/16/moonshots-upcoming-kimi-3…
@ThinkingMachines releases Inkling, a 975B open-weight multimodal MoE with 41B active parameters.
marktechpost.com/2026/07/15/thinking-machines-lab-r…
PrismML's Bonsai 27B model compresses to under 4GB, running on an iPhone with 90% of its original performance.
the-decoder.com/bonsai-27b-is-a-full-open-reasoning…
PrismML releases Bonsai 27B, a 1-bit and ternary build of Qwen3.6-27B running on laptops and phones.
marktechpost.com/2026/07/14/prismml-releases-bonsai…
@AnthropicAI's Claude Sonnet 5 narrows agentic coding gap to Opus 4.8 at lower token pricing.
marktechpost.com/2026/07/13/anthropic-claude-sonnet…
@AnthropicAI's Claude Code gains a built-in browser for direct web interaction inside the dev environment.
the-decoder.com/claude-code-now-has-a-built-in-brow…
@KyutAI releases MuScriptor, an open-weight Transformer transcribing full multi-instrument mixes to MIDI.
marktechpost.com/2026/07/10/kyutai-releases-muscrip…
@OpenAI's GPT-5.6 Sol autonomously post-trained the smaller Luna model with a "fairly underspecified prompt".
the-decoder.com/openais-gpt-5-6-sol-autonomously-po…
OpenAI staffer Vaibhav Srivastav maps GPT-5.6 Sol's five reasoning levels to task complexity.
the-decoder.com/openai-staffer-maps-out-which-of-gp…
OpenAI releases GPT-5.6 with three model tiers and programmatic tool calling.
marktechpost.com/2026/07/09/openai-releases-gpt-5-6…
@Meta releases Muse Spark 1.1, a multimodal reasoning model for agentic tasks on its Meta Model API.
marktechpost.com/2026/07/09/meta-superintelligence-…
OpenAI's GPT-Live enables simultaneous listening and speaking in ChatGPT using full-duplex architecture.
the-decoder.com/chatgpt-can-now-listen-and-talk-at-…
OpenAI launches GPT-5.6 after government approval, ending limited preview for approved organizations. Announces 'ChatGPT Work' for enterprise use. Source: theverge.com/ai-artificial-intelligence/963464/open…
@MistralAI enters robotics with Robostral Navigate, an 8B model steering robots using just one camera.
the-decoder.com/mistral-enters-robotics-with-robost…
New AI model uses deep reasoning to understand structure-property relationships across biology, chemistry and materials science.
arxiv.org/abs/2607.07708v1
@Robbyant open-sourced LingBot-Vision, a 1B vision model using masked boundary modeling for dense spatial perception.
marktechpost.com/2026/07/07/ant-groups-robbyant-ope…
@NVIDIA releases Audex, a unified audio-text LLM preserving text intelligence from its Nemotron-Cascade-2 backbone.
marktechpost.com/2026/07/07/nvidia-releases-audex-n…
@Cohere releases Transcribe Arabic, an open-source model outperforming Whisper on Arabic dialects and code-switching.
the-decoder.com/cohere-transcribe-arabic-is-an-open…
Zhipu AI launches ZCode with GLM-5.2 to challenge Claude Code and OpenAI Codex at a fraction of the cost.
the-decoder.com/zhipu-ai-launches-zcode-to-challeng…
Anthropic’s Claude Cowork now on web and mobile for Max subscribers.
@Tencent releases Hy3, a 295B parameter open-source model using 21B active parameters to match larger models.
the-decoder.com/tencent-releases-hy3-open-source-mo…
@Meituan releases LongCat-2.0, a 1.6T-parameter MoE model with 1M token context and 48B active parameters per token.
marktechpost.com/2026/07/05/meituan-releases-longca…
Baidu's Unlimited OCR processes dozens of document pages in one pass using a memory trick inspired by human forgetting.
the-decoder.com/baidus-unlimited-ocr-processes-doze…
@AnthropicAI developer Thariq Shihipar says with Claude 3.5 Sonnet, your blind spots—not the model—are the bottleneck. the-decoder.com/anthropic-developer-shares-promptin…
@MistralAI releases Leanstral 1.5, a 119B mixture-of-experts model solving 587 of 672 PutnamBench problems.
marktechpost.com/2026/07/03/mistral-ai-releases-lea…
Search agents using LLMs now clarify ambiguous queries before deep reasoning to solve complex tasks.
huggingface.co/papers/2606.27669
@AnthropicAI's Claude Sonnet 5 edges past Opus 4.8 on the GDPval-AA v2 knowledge test with a score of 1,618.
the-decoder.com/anthropics-new-claude-sonnet-5-clos…
@Meta AI releases Brain2Qwerty v2, a non-invasive MEG brain-to-text pipeline decoding typed sentences at 61% word accuracy.
marktechpost.com/2026/06/30/meta-ai-releases-brain2…
OpenAI tests gifting Codex credits as a new growth strategy.
testingcatalog.com/openai-tests-gifting-codex-credi…
OpenAI is testing Office-specific controls in Codex for PowerPoint and Excel add-ins.
testingcatalog.com/openai-prepares-upgraded-excel-a…
Only three AI models survived 500 days running a simulated software company without going broke.
the-decoder.com/only-three-ai-models-finished-above…
@DeepSeek_AI released DSpark, a speculative decoding framework that accelerates DeepSeek-V4 per-user generation by 60–85% over MTP-1.
marktechpost.com/2026/06/27/deepseek-releases-dspar…
@Google Research's Gemini-SQL2 tops text-to-SQL benchmarks at 80.04% accuracy, outperforming OpenAI and Anthropic.
huggingface.co/papers/2606.27277
@AnthropicAI's Claude Opus 4.7 solves 56% of a single MirrorCode task in 19 days at a cost of $2,600.
the-decoder.com/an-ai-model-programmed-nonstop-for-…
Gradium launches stt-translate and s2s-translate, real-time speech translation models outperforming gpt-realtime-translate on accuracy and latency.
marktechpost.com/2026/06/24/gradium-launches-stt-tr…
OpenAI improves ChatGPT Instant with better intent recognition and multi-turn context.
the-decoder.com/openai-says-chatgpt-instant-now-bet…
OTHER CATEGORIES