AI & LLMs
EmbeddingGemma 2 puts multimodal search on a phone
Google DeepMind's EmbeddingGemma 2 maps text, code, images, video and audio into one vector space with 740M parameters. Text-only use needs about 191 MB of RAM.
Open Weights news and analysis
Open-weight model releases, meaning models whose weights can be downloaded and run outside the vendor's cloud: GLM, Qwen and Granite launches, the licenses attached to them, and tools like Ollama that run them locally. Each story states the parameter count, license terms and hardware needed.
AI & LLMs
Strands Decider 2B makes agent choices in 115 ms
AWS's Strands Decider 2B is an open-source model that picks from a list instead of writing text. It answers in 115 ms on an RTX 3090 and scores 72.3% on JevBench.
AI & LLMs
Mistral Large 4 launches with 1.05T parameters
Mistral Large 4 is a 1.05T-parameter open-weight MoE model with a 1M-token context, in API preview at $1.36 per million input tokens. Weights follow in October.
AI & LLMs
Reflection AI launches Beam, a 501B open-weight model
Reflection AI's Beam has 501B parameters but uses 23B per token, scores 80.9 on SWE-Bench Verified, and ships weights under Apache 2.0 later in October.
AI & LLMs
Cloudflare Clef: open-weight models that return decisions
Cloudflare's Clef (27B) and Clef-flash (9B) answer typed questions instead of writing text, at 209 ms and 39 ms median latency, under Apache 2.0.
AI & LLMs
Aleph Alpha Kolibri 1: open German-English MoE model
Aleph Alpha released Kolibri 1 on October 3: a 78.1B-parameter open model with 3.46B active, built for German and English, under the Apache 2.0 license.
AI & LLMs
AstaBrief 8B: Ai2's open model writes cited reports
Ai2 released AstaBrief 8B, an Apache 2.0 model on Qwen3-8B that writes cited research reports in 51.1 seconds, 3.5x faster than Asta's Thinking mode.
AI & LLMs
Qwen's research license, explained
The Qwen Research License lets anyone download Qwen-Image-2.1 but limits it to research or evaluation. Commercial use needs a separate license from Qwen.
AI & LLMs
Xiaomi MiMo-V2.6-Pro ships 1T open weights under MIT
Xiaomi's MiMo-V2.6-Pro has 1.02 trillion parameters, 42 billion active, a 1M-token context and MIT-licensed weights, but needs about 680 GB of GPU memory.
AI & LLMs
NVIDIA Kumo Tabular tops TabArena with open weights
NVIDIA's Kumo Tabular predicts from a labeled table in one pass, with no training run. It ranks first on TabArena at 1,950 Elo, in sizes from 28M to 215M.
AI & LLMs
Holo4 open-weight agent models score 61.7% on OSWorld 2.0
H Company's Holo4-27B scores 61.7% on OSWorld 2.0 at $1.22 a task, but only its Apache 2.0 sibling, 35B-A3B, can be self-hosted commercially.
AI & LLMs
Laya's typed decisions, explained
A typed-decision model fills in fixed questions with typed answers and a confidence score. Laya's 421M open weights let you verify that yourself.
AI & LLMs
Liquid AI LFM2.5-VL-DSpark speeds up its 3B vision model
Liquid AI's 280M-parameter LFM2.5-VL-DSpark draft model decodes its 3B vision model up to 3.13x faster on an M5 Max, with identical output.
AI & LLMs
Qwen-Image-2.1 ships 7B weights under a research license
Qwen-Image-2.1 puts 7 billion parameters and native transparency in a downloadable image model. Its license bars commercial use without permission.
AI & LLMs
Laya is a 421M open-weights answer to Jev
Laya returns typed decisions in 32.8 milliseconds from 421M parameters under Apache 2.0. Its zero-shot accuracy is close to chance.
AI & LLMs
Salesforce Koa builds on open weights but ships closed
Salesforce says Koa makes 3x fewer errors on its own CRM Bench, but the weights are proprietary and the model runs only on Salesforce infrastructure.
AI & LLMs
Qwen3.5-9B spends 32KB per token on KV cache
Qwen3.5-9B runs full attention on only 8 of its 32 layers, so its KV cache costs 32KB per token. That ratio decides whether it fits in 8GB.
Startup Watch
Mistral raises €3B at €21B, led by Samsung
Mistral raised €3 billion at a €21 billion valuation, nearly double its last round. It plans 1 GW of European compute by 2030 and will host rivals' open weights.
AI & LLMs
GLM-5.3 weights ship under a custom license and need eight GPUs
Z.ai's flagship GLM-5.3 weights are on Hugging Face after a safety review delay. The license is not MIT, and the model needs at least eight high-end GPUs.
AI & LLMs
Qwen3.8-Flash-Next holds 125B parameters but runs only 6B
Alibaba released Qwen3.8-Flash-Next on August 26, 2026. It activates 6B of 125B parameters per token, costs $0.16 per million input tokens, and previews Qwen4.
Dev Stack
Ollama 0.33 lets Claude Desktop run Qwen, DeepSeek and Kimi locally
Ollama 0.33.0 adds a local proxy that swaps the model behind Anthropic's Claude Desktop app for an open one. A spring attempt failed on model-ID checks.
AI & LLMs
Z.ai and IBM ship open-weight reasoning models a day apart
Z.ai released GLM-5.3-Flash under MIT with 320B parameters. IBM released Granite 4.2 under Apache 2.0 at 3B to 30B. They suit very different hardware.
The weekly digest
One email every Friday with the week's top stories from all six desks: AI, dev tools, coding, the tech industry, startups and what's next. Free, no spam.