Muse Code Explained: Meta's Multi-Agent CLI (2026)
layer3labs.ioA technical explainer on Muse Code, Meta's CLI coding agent released Aug 5, 2026, backed by Muse Spark 1.2, covering its multi-agent architecture, event-log auditability, persistent memory, and pricing tiers.
Meta's Muse Spark: a smaller, faster AI model for broad app deployment
infoworld.comThe first model from Meta Superintelligence Labs, Muse Spark reflects a shift toward efficient, product-ready AI as enterprises weigh cost, latency, and real-world deployment.
HUMAIN Unveils humain-m3, a Frontier Arabic Language Model Developed by MiniMax
prnewswire.comHUMAIN announced humain-m3, a 428-billion parameter MoE model built on the MiniMax-M3 lineage and pre-trained on over one trillion Arabic-native tokens, achieving the highest average score among frontier models across seven Arabic benchmarks.
Saudi Firm Humain Builds 428B Arabic LLM on MiniMax
shattered.ioAnalysis of humain-m3, a 428B parameter Arabic LLM built on MiniMax-M3 with further pre/post-training on over 1 trillion Arabic-native tokens, available in research preview via HUMAIN Node.
HUMAIN unveils 428B parameter Arabic model with MiniMax
middleeastainews.comSaudi Arabia's national AI company HUMAIN introduced humain-m3, a 428-billion parameter Arabic mixture-of-experts model built on the MiniMax-M3 lineage, pre-trained on over one trillion Arabic-native tokens, scoring highest on seven public Arabic benchmarks.
LFM2.5-DSpark: Up to 3.2x Faster Inference from H100 to MacBook
liquid.aiLiquid AI releases DSpark draft model checkpoints for three LFM2.5 models, adding speculative decoding paths that reach up to 3.18x throughput on GPU and 2.87x on-device without changing output quality.
Liquid AI released LFM2.5-VL-3B-DSpark
datanorth.aiDataNorth reports Liquid AI released LFM2.5-VL-3B-DSpark, a 279.5M-parameter open-weight speculative-decoding drafter for its 3B vision-language model, accelerating decoding up to 3.13x on edge hardware.
LFM2.5-VL-DSpark: Accelerating vision-language models on edge and beyond
liquid.aiLiquid AI releases an experimental DSpark draft model for its LFM2.5-VL-3B vision-language model, using speculative decoding to deliver up to 3.13x faster decoding on GPUs and 3.13x on edge devices without changing output quality.
Introducing LFM2: The Fastest On-Device Foundation Models on the Market
liquid.aiLiquid AI introduces LFM2, a new class of Liquid Foundation Models built on a hybrid architecture delivering 2x faster decode and prefill performance than Qwen3 on CPU for on-device AI agents.
Liquid AI AI News - Latest Updates, Announcements & Releases
pricepertoken.comPrice per token tracker of Liquid AI LFM releases, including LFM2.5-2.6B (open source, released Aug 2026) and other edge-focused models with pricing and availability details.
News — Liquid AI
liquid.aiLiquid AI's news page listing recent LFM model releases and engineering updates, including LFM2.5-VL-DSpark, MacPaw partnership for on-device AI, and LFM2.5-2.6B for edge agents.
MiniMax raises $619M in Hong Kong IPO as investor appetite for generative AI remains strong
siliconangle.comChinese AI startup MiniMax Group went public in Hong Kong in January 2026, raising about US$619 million at the top of its price range, with an oversubscribed offering.
MiniMax Launches humain-m3, a 428 Billion Parameter Arabic Model
sacra.comMiniMax launched humain-m3, a 428 billion parameter Arabic language model setting new standards in Arabic processing, per Sacra's MiniMax profile covering revenue, funding, and recent developments.
MiniMax Announces First Half 2026 Financial Results
prnewswire.comMiniMax reported H1 2026 revenue up 283.1% to US$116.6 million, with AI-native product revenue up 100.9%; R&D spending rose 138.8% on cloud training costs.
MiniMax M3 Release Date And The Sparse Attention Upgrade
linkedin.comAnalysis of MiniMax M3's shift from full attention to MiniMax Sparse Attention, with tiled and I/O improvements for long-context inference efficiency.
MiniMax M3 - API Pricing & Benchmarks
openrouter.aiOpenRouter page for MiniMax M3, released May 31, 2026, a native multimodal model with 1M-token context, function calling, and structured outputs, served by 13 providers.
MiniMax M3: Release Date, API, Pricing, and AI Agent Use Cases
myclaw.aiGuide to MiniMax M3, a native multimodal LLM with up to 1M-token context, MiniMax Sparse Attention, tool use, coding, and agentic workflows, available via API.
Chinese AI Firm MiniMax's ARR Jumps to $800 Million
app.dealroom.coMiniMax's annualized revenue run rate jumped to US$800 million from US$150 million in February, driven by model inference demand, though losses more than doubled.
China's MiniMax sees revenue nearly quadruple in first half as AI demand surges
investing.comChinese AI startup MiniMax reported first-half 2026 revenue up 283% year-on-year to US$116.6 million, driven by surging demand for its low-cost models and AI inference platform.
Meta is licensing Midjourney's AI image and video tech
engadget.comEngadget covers Meta's partnership with Midjourney, in which Meta will license the startup's aesthetic technology for future models and products, bringing beauty to billions of users.