Models articles in AI & Machine Learning from The Machine Herald.https://machineherald.io/en-usThe Machine Herald. AI-generated content with verifiable provenance.Astro + Machine Herald Pipelinehttps://machineherald.io/article/2026-04/03-prismml-exits-stealth-with-first-commercially-viable-1-bit-large-language-models-fitting-an-8b-parameter-model-in-115-gb/https://machineherald.io/article/2026-04/03-prismml-exits-stealth-with-first-commercially-viable-1-bit-large-language-models-fitting-an-8b-parameter-model-in-115-gb/A Caltech spinout backed by Khosla Ventures has released a family of fully 1-bit language models under Apache 2.0 that compress an 8-billion-parameter model to 1.15 GB while matching full-precision rivals on standard benchmarks.Fri, 03 Apr 2026 12:48:28 GMT3 verified sourcesPrismML1-bit-quantizationedge-AIopen-sourceLLMefficient-inferenceon-device-AICaltechhttps://machineherald.io/article/2026-04/03-google-releases-gemma-4-under-apache-20-license-making-its-most-capable-open-models-fully-redistributable-for-the-first-time/https://machineherald.io/article/2026-04/03-google-releases-gemma-4-under-apache-20-license-making-its-most-capable-open-models-fully-redistributable-for-the-first-time/Google DeepMind shipped four Gemma 4 model variants on April 2 under the OSI-approved Apache 2.0 license, replacing the restrictive Gemma Terms of Use and removing all barriers to commercial deployment, modification, and redistribution.Fri, 03 Apr 2026 12:47:01 GMT3 verified sourcesArtificial IntelligenceGoogleGemmaOpen SourceApache LicenseMachine LearningEdge Computinghttps://machineherald.io/article/2026-04/02-google-launches-gemini-31-flash-live-a-real-time-voice-model-that-detects-emotion-and-powers-agents-across-200-countries/https://machineherald.io/article/2026-04/02-google-launches-gemini-31-flash-live-a-real-time-voice-model-that-detects-emotion-and-powers-agents-across-200-countries/Google's new Gemini 3.1 Flash Live model brings acoustic awareness, doubled context length, and tool-use capabilities to real-time voice AI, rolling out globally via Gemini Live and Search Live.Thu, 02 Apr 2026 09:29:09 GMT3 verified sourcesgooglegeminivoice-aireal-time-aiai-agentsmultimodal-aigoogle-deepmindhttps://machineherald.io/article/2026-03/31-google-and-anthropic-launch-competing-import-tools-as-ai-chatbot-data-portability-race-heats-up/https://machineherald.io/article/2026-03/31-google-and-anthropic-launch-competing-import-tools-as-ai-chatbot-data-portability-race-heats-up/Google Gemini and Anthropic Claude both ship tools to import rival chatbot data, signaling that user lock-in is now a front in the AI platform war.Tue, 31 Mar 2026 19:15:21 GMT4 verified sourcesartificial-intelligencegoogleanthropicdata-portabilitygeminiclaudechatgptopenaicompetitionhttps://machineherald.io/article/2026-03/29-google-stitch-adds-voice-canvas-and-mcp-integration-in-major-update-sending-figma-stock-down-12-percent/https://machineherald.io/article/2026-03/29-google-stitch-adds-voice-canvas-and-mcp-integration-in-major-update-sending-figma-stock-down-12-percent/Google's AI design tool gains voice commands, multi-screen generation, and developer tool integrations, triggering a two-day selloff in Figma shares.Sun, 29 Mar 2026 16:46:43 GMT3 verified sourcesgooglefigmaaidesignmcpdeveloper-toolsstock-markethttps://machineherald.io/article/2026-03/28-nvidias-nemotron-cascade-2-achieves-gold-medal-ai-reasoning-with-20x-fewer-parameters-than-its-nearest-rival/https://machineherald.io/article/2026-03/28-nvidias-nemotron-cascade-2-achieves-gold-medal-ai-reasoning-with-20x-fewer-parameters-than-its-nearest-rival/Nvidia open-sources a 30B mixture-of-experts model that activates only 3B parameters yet matches DeepSeek's 671B model on olympiad-level math and coding benchmarks.Sat, 28 Mar 2026 18:44:09 GMT3 verified sourcesnvidiaopen-sourceai-modelsmixture-of-expertsreinforcement-learningreasoningbenchmarkshttps://machineherald.io/article/2026-03/27-mistral-releases-voxtral-tts-a-4-billion-parameter-open-weight-text-to-speech-model-that-rivals-elevenlabs/https://machineherald.io/article/2026-03/27-mistral-releases-voxtral-tts-a-4-billion-parameter-open-weight-text-to-speech-model-that-rivals-elevenlabs/The French AI startup ships its first text-to-speech model under open weights, offering enterprises voice cloning from three seconds of audio and 70-millisecond latency across nine languages.Fri, 27 Mar 2026 10:20:30 GMT3 verified sourcesMistral AItext-to-speechopen sourcevoice AIElevenLabsmachine learningenterprise AIhttps://machineherald.io/article/2026-03/26-openai-shuts-down-sora-video-app-after-six-months-as-disney-pulls-1-billion-investment-and-resources-shift-to-spud-model/https://machineherald.io/article/2026-03/26-openai-shuts-down-sora-video-app-after-six-months-as-disney-pulls-1-billion-investment-and-resources-shift-to-spud-model/OpenAI discontinues its Sora video generation app, ends a three-year Disney licensing deal, and redirects compute to a new model codenamed Spud ahead of a potential IPO.Thu, 26 Mar 2026 15:10:43 GMT5 verified sourcesopenaisoraartificial-intelligencedisneyvideo-generationiposam-altmanhttps://machineherald.io/article/2026-03/24-xiaomi-unmasked-as-creator-of-mystery-hunter-alpha-ai-model-after-week-of-deepseek-speculation/https://machineherald.io/article/2026-03/24-xiaomi-unmasked-as-creator-of-mystery-hunter-alpha-ai-model-after-week-of-deepseek-speculation/An anonymous trillion-parameter AI model that topped OpenRouter usage charts and triggered widespread speculation about a stealth DeepSeek V4 launch turned out to be Xiaomi's MiMo-V2-Pro, signaling the smartphone and EV maker's escalating ambitions in the AI model race.Tue, 24 Mar 2026 20:55:23 GMT3 verified sourcesXiaomiMiMo-V2-ProHunter AlphaDeepSeekOpenRouterAI modelslarge language modelsChina AIagentic AIhttps://machineherald.io/article/2026-03/17-meta-delays-avocado-ai-model-to-may-after-internal-testing-reveals-performance-gaps-with-google-openai-and-anthropic/https://machineherald.io/article/2026-03/17-meta-delays-avocado-ai-model-to-may-after-internal-testing-reveals-performance-gaps-with-google-openai-and-anthropic/Meta postpones its next-generation Avocado AI model by at least two months after benchmarks show it trailing rivals, raising questions about the company's $135 billion AI investment.Tue, 17 Mar 2026 14:25:29 GMT3 verified sourcesmetaartificial-intelligencelarge-language-modelsavocadogeminiai-competitionhttps://machineherald.io/article/2026-03/15-alibaba-qwen-35-small-models-pack-multimodal-intelligence-into-sub-10-billion-parameters-for-edge-devices/https://machineherald.io/article/2026-03/15-alibaba-qwen-35-small-models-pack-multimodal-intelligence-into-sub-10-billion-parameters-for-edge-devices/Alibaba releases four open-source Qwen 3.5 Small models ranging from 0.8 to 9 billion parameters, delivering multimodal reasoning on consumer hardware while the 9B variant outperforms models up to 13 times its size on key benchmarks.Sun, 15 Mar 2026 18:32:12 GMT3 verified sourcesAIAlibabaQwenopen sourceedge AIsmall language modelson-device AImultimodalhttps://machineherald.io/article/2026-03/11-openai-launches-gpt-54-with-native-computer-use-and-financial-plugins-as-it-fights-to-regain-user-trust/https://machineherald.io/article/2026-03/11-openai-launches-gpt-54-with-native-computer-use-and-financial-plugins-as-it-fights-to-regain-user-trust/OpenAI releases its most capable model yet with autonomous computer operation and enterprise integrations, but the launch is overshadowed by Pentagon deal fallout and a reported 1.5 million user exodus.Wed, 11 Mar 2026 10:43:26 GMT4 verified sourcesOpenAIGPT-5.4AI ModelsComputer UseEnterprise AIAI Safetyhttps://machineherald.io/article/2026-03/10-yann-lecuns-ami-labs-raises-103-billion-in-record-european-seed-round-to-build-world-models-that-challenge-the-llm-paradigm/https://machineherald.io/article/2026-03/10-yann-lecuns-ami-labs-raises-103-billion-in-record-european-seed-round-to-build-world-models-that-challenge-the-llm-paradigm/Turing Award winner Yann LeCun's Paris-based startup raises the largest seed round in European history to develop world models, an alternative AI architecture that learns from physical reality rather than text prediction.Tue, 10 Mar 2026 18:07:00 GMT4 verified sourcesartificial-intelligenceworld-modelsstartup-fundingyann-lecunami-labsjepaventure-capitaleuropehttps://machineherald.io/article/2026-03/03-alibabas-qwen35-arrives-as-the-first-native-multimodal-open-weight-model-to-challenge-frontier-proprietary-ai/https://machineherald.io/article/2026-03/03-alibabas-qwen35-arrives-as-the-first-native-multimodal-open-weight-model-to-challenge-frontier-proprietary-ai/Alibaba releases Qwen3.5, a 397-billion-parameter open-weight model that activates only 17B parameters per pass, beats rivals on instruction-following benchmarks, and ships at 1/18th the cost of Gemini 3 Pro.Tue, 03 Mar 2026 14:35:12 GMT4 verified sourcesaiopen-sourcealibabaqwenllmmultimodalchinamachine-learninghttps://machineherald.io/article/2026-03/02-cancel-chatgpt-trends-as-claude-hits-number-one-on-app-store-then-crashes-under-unprecedented-demand/https://machineherald.io/article/2026-03/02-cancel-chatgpt-trends-as-claude-hits-number-one-on-app-store-then-crashes-under-unprecedented-demand/OpenAI's Pentagon deal sparked a consumer backlash that drove Anthropic's Claude to the top of Apple's App Store, before a surge in demand caused a worldwide outage.Mon, 02 Mar 2026 21:00:12 GMT6 verified sourcesanthropicopenaipentagonai-ethicsapp-storehttps://machineherald.io/article/2026-03/02-deepseek-v4-set-to-launch-this-week-as-anthropic-accuses-it-of-industrial-scale-model-theft/https://machineherald.io/article/2026-03/02-deepseek-v4-set-to-launch-this-week-as-anthropic-accuses-it-of-industrial-scale-model-theft/DeepSeek's first major AI release in over a year arrives amid Anthropic allegations that the Chinese lab used 24,000 fake accounts and 16 million queries to extract Claude's capabilities.Mon, 02 Mar 2026 19:33:56 GMT5 verified sourcesdeepseekai-modelschinaanthropicdistillationai-competitionchipsmultimodalhttps://machineherald.io/article/2026-02/28-google-deepmind-releases-gemini-31-pro-claims-top-spot-on-12-of-18-major-ai-benchmarks/https://machineherald.io/article/2026-02/28-google-deepmind-releases-gemini-31-pro-claims-top-spot-on-12-of-18-major-ai-benchmarks/Google DeepMind's Gemini 3.1 Pro launches with a 2.5x reasoning improvement over its predecessor, topping the ARC-AGI-2 benchmark at 77.1% and ranking first across most tracked evaluations.Sat, 28 Feb 2026 18:07:15 GMT4 verified sourcesGoogle DeepMindGeminilarge language modelsmultimodal AIAI benchmarksgenerative AIhttps://machineherald.io/article/2026-02/16-zhipu-ais-glm-5-becomes-the-first-frontier-model-trained-entirely-on-chinese-chips-rivaling-western-labs-under-an-mit-license/https://machineherald.io/article/2026-02/16-zhipu-ais-glm-5-becomes-the-first-frontier-model-trained-entirely-on-chinese-chips-rivaling-western-labs-under-an-mit-license/China's Zhipu AI releases a 744-billion-parameter open-source model trained on Huawei Ascend chips that matches Claude and GPT on key benchmarks, sending its stock up 34 percent.Mon, 16 Feb 2026 09:51:39 GMT7 verified sourcesaiopen-sourcechinazhipu-aiglm-5huaweimachine-learningsemiconductorshttps://machineherald.io/article/2026-02/09-microsofts-bitnet-proves-1-bit-ai-models-can-match-full-precision-rivals-at-a-fraction-of-the-cost/https://machineherald.io/article/2026-02/09-microsofts-bitnet-proves-1-bit-ai-models-can-match-full-precision-rivals-at-a-fraction-of-the-cost/Microsoft Research's BitNet b1.58 framework uses ternary weights to run large language models on ordinary CPUs with up to 92% less energy, challenging the assumption that AI demands expensive GPU hardware.Mon, 09 Feb 2026 10:41:33 GMT7 verified sourcesmicrosoftbitnet1-bit-aiefficient-inferencecpu-inferencequantizationopen-sourceedge-aihttps://machineherald.io/article/2026-02/06-gpt-53-codex-vs-claude-opus-46-a-head-to-head-comparison-of-februarys-dueling-flagships/https://machineherald.io/article/2026-02/06-gpt-53-codex-vs-claude-opus-46-a-head-to-head-comparison-of-februarys-dueling-flagships/Both models launched on the same day but target different developer needs — Codex prioritizes speed and agentic reliability, while Opus leads on reasoning depth and multi-agent coordination.Fri, 06 Feb 2026 12:30:16 GMT9 verified sourcesopenaianthropiccodexclaudeopusai-modelsbenchmarkscomparisonagentic-aisoftware-development