---
title: "AlphaSignal Llms: Mixture Of Experts News"
url: "https://alphasignal.ai/topic/llms/mixture-of-experts"
description: "Latest AI and machine learning news about Mixture Of Experts (Llms) on AlphaSignal."
---

# AlphaSignal Llms: Mixture Of Experts News

Latest AI and machine learning news about Mixture Of Experts (Llms) on AlphaSignal.

## Articles

- [Aleph Alpha's Kolibri-1 Brings Open German-English Reasoning to a Single GPU](https://alphasignal.ai/news/aleph-alpha-s-kolibri-1-brings-open-german-english-reasoning-to-a-single-gpu.md) (2026-10-07)
- [Mistral AI's Mistral Large 4 Packs 1 Trillion Parameters Into 4 GPUs](https://alphasignal.ai/news/mistral-ai-s-mistral-large-4-packs-1-trillion-parameters-into-4-gpus.md) (2026-10-06)
- [Ant Group's Ling 3.1 Flash Doubles Agent Scores With 560B Parameters](https://alphasignal.ai/news/ant-group-s-ling-3-1-flash-doubles-agent-scores-with-560b-parameters.md) (2026-10-06)
- [Reflection AI's Beam Challenges DeepSeek With 501B Open-Weight Reasoning Model](https://alphasignal.ai/news/reflection-ai-s-beam-challenges-deepseek-with-501b-open-weight-reasoning-model.md) (2026-10-05)
- [Mia's AI Lab Ships GLM-5.3-Flash With Uncensored Weights and 16x Faster Responses](https://alphasignal.ai/news/mia-s-ai-lab-ships-glm-5-3-flash-with-uncensored-weights-and-16x-faster.md) (2026-10-05)
- [China Telecom's Xing4.0 Runs a 29B Coding Agent on 19GB Locally](https://alphasignal.ai/news/china-telecom-s-xing4-0-runs-a-29b-coding-agent-on-19gb-locally.md) (2026-10-04)
- [Aleph Alpha Releases Kolibri-1, an Open German Reasoning Model With 1M-Token Context](https://alphasignal.ai/news/aleph-alpha-releases-kolibri-1-an-open-german-reasoning-model-with-1m-token.md) (2026-10-04)
- [LOOM Pushes MoE Transformers to 12 Loops, Boosting Accuracy by 5.3 Points](https://alphasignal.ai/news/loom-pushes-moe-transformers-to-12-loops-boosting-accuracy-by-5-3-points.md) (2026-10-01)
- [Upstage's Solar Mini 4 Triples Its Benchmark Score With 3B Active Parameters](https://alphasignal.ai/news/upstage-s-solar-mini-4-triples-its-benchmark-score-with-3b-active-parameters.md) (2026-10-01)
- [ID Balancing Cuts Routing Overload by 89% in 70B Sparse AI Models](https://alphasignal.ai/news/id-balancing-cuts-routing-overload-by-89-in-70b-sparse-ai-models.md) (2026-09-30)
- [Stanford's SOMA Slashes AI Training Noise by Sharding Models Into Expert Shards](https://alphasignal.ai/news/stanford-s-soma-slashes-ai-training-noise-by-sharding-models-into-expert-shards.md) (2026-09-29)
- [Blackfrost-AI's CYBER-FROST Brings a 180B Security Model to Local Machines](https://alphasignal.ai/news/blackfrost-ai-s-cyber-frost-brings-a-180b-security-model-to-local-machines.md) (2026-09-27)
- [Yandex Drops AliceAI Foundation 80B Open Source Model That Beats Rivals on Math](https://alphasignal.ai/news/yandex-drops-aliceai-foundation-80b-open-source-model-that-beats-rivals-on-math.md) (2026-09-23)
- [HySparse2 Slashes AI Agent Memory 4.5x While Cutting Prefill Work in Half](https://alphasignal.ai/news/hysparse2-slashes-ai-agent-memory-4-5x-while-cutting-prefill-work-in-half.md) (2026-09-22)
- [StepFun's Step 5 Preview Matches Top Rivals at One-Third the Cost](https://alphasignal.ai/news/stepfun-s-step-5-preview-matches-top-rivals-at-one-third-the-cost.md) (2026-09-22)
- [Xiaomi's MiMo-V2.6-Pro Tops Open-Weight Rankings With a 1T-Parameter Model](https://alphasignal.ai/news/xiaomi-s-mimo-v2-6-pro-tops-open-weight-rankings-with-a-1t-parameter-model.md) (2026-09-21)
- [Invergent's Open-Weight Rune v3 Hits Second on the Decision Index](https://alphasignal.ai/news/invergent-s-open-weight-rune-v3-hits-second-on-the-decision-index.md) (2026-09-21)
- [NVIDIA Releases DeepSeek-V4.1-Flash Checkpoint Optimized for Blackwell GPUs](https://alphasignal.ai/news/nvidia-releases-deepseek-v4-1-flash-checkpoint-optimized-for-blackwell-gpus.md) (2026-09-16)
- [A 7.4B Model Matches GPT-3 13B Using 20x Less Compute](https://alphasignal.ai/news/a-7-4b-model-matches-gpt-3-13b-using-20x-less-compute.md) (2026-09-16)
- [Ant Group Releases Ling-3.0-flash-Fin to Self-Host Financial Research at 124B Parameters](https://alphasignal.ai/news/ant-group-releases-ling-3-0-flash-fin-to-self-host-financial-research-at-124b.md) (2026-09-16)
- [China Telecom Releases Xing4.0, an Open 29B Coding Agent Built on Huawei Chips](https://alphasignal.ai/news/china-telecom-releases-xing4-0-an-open-29b-coding-agent-built-on-huawei-chips.md) (2026-09-16)
- [IFM's K2-Horizon 7B Beats GPT-5 on Coding With 70.6 SWE-bench Score](https://alphasignal.ai/news/ifm-s-k2-horizon-7b-beats-gpt-5-on-coding-with-70-6-swe-bench-score.md) (2026-09-15)
- [What DeepSeek-V4.1-Flash teaches us about efficient AI](https://alphasignal.ai/news/what-deepseek-v4-1-flash-teaches-us-about-efficient-ai.md) (2026-09-14)
- [Edge0 Streams an 8B AI Model From SSD Using Only 1 GiB](https://alphasignal.ai/news/edge0-streams-an-8b-ai-model-from-ssd-using-only-1-gib.md) (2026-09-13)
- [Edge0 Runs a 35B AI Model on an iPhone With Just 2.9 GiB](https://alphasignal.ai/news/edge0-runs-a-35b-ai-model-on-an-iphone-with-just-2-9-gib.md) (2026-09-12)
- [Ant Group's Ling-3.0-flash-VL Adds Vision to a 124B Open Reasoning Model](https://alphasignal.ai/news/ant-group-s-ling-3-0-flash-vl-adds-vision-to-a-124b-open-reasoning-model.md) (2026-09-11)
- [DeepSeek's V4.1-Flash Beats Its Own Flagship at a Quarter of the Memory Cost](https://alphasignal.ai/news/deepseek-s-v4-1-flash-beats-its-own-flagship-at-a-quarter-of-the-memory-cost.md) (2026-09-10)
- [Aikido Security's Altar-1 Shrinks a 1.5TB Security AI by 78%](https://alphasignal.ai/news/aikido-security-s-altar-1-shrinks-a-1-5tb-security-ai-by-78.md) (2026-09-07)
- [IST Austria Shrinks a 177B Qwen3.8-Flash-Next Model to 66 GB](https://alphasignal.ai/news/ist-austria-shrinks-a-177b-qwen3-8-flash-next-model-to-66-gb.md) (2026-09-07)
- [Ant Group's Ling-3.0-flash-VL Brings Images and Video to Its 124B Open Model](https://alphasignal.ai/news/ant-group-s-ling-3-0-flash-vl-brings-images-and-video-to-its-124b-open-model.md) (2026-09-04)
- [Tsinghua's SMELT Cuts AI Training Costs 18% by Looping Layers Twice](https://alphasignal.ai/news/tsinghua-s-smelt-cuts-ai-training-costs-18-by-looping-layers-twice.md) (2026-09-01)
- [PeasantSmith Squeezes Alibaba's 177B Qwen3.8-Flash-Next Into a 75 GB Local File](https://alphasignal.ai/news/peasantsmith-squeezes-alibaba-s-177b-qwen3-8-flash-next-into-a-75-gb-local-file.md) (2026-08-31)
- [DeepSeek's V4-Flash-Vision-Exp Adds Image Understanding at Half the Cost of Rivals](https://alphasignal.ai/news/deepseek-s-v4-flash-vision-exp-adds-image-understanding-at-half-the-cost-of.md) (2026-08-31)
- [OrcaRouter Strips Safety Guardrails From Z.ai's 320B GLM-5.3-Flash](https://alphasignal.ai/news/orcarouter-strips-safety-guardrails-from-z-ai-s-320b-glm-5-3-flash.md) (2026-08-29)
- [Alibaba's Qwen3.8-Flash-Next Hits 10x Long-Context Speed at 1/9 Training Cost](https://alphasignal.ai/news/alibaba-s-qwen3-8-flash-next-hits-10x-long-context-speed-at-1-9-training-cost.md) (2026-08-26)
- [Z.ai's GLM-5.3 Got So Dangerous It Delayed Its Own Release](https://alphasignal.ai/news/z-ai-s-glm-5-3-got-so-dangerous-it-delayed-its-own-release.md) (2026-08-25)
- [Daily's PhoneLLM Matches GPT Quality at 94% Lower Cost for Voice Agents](https://alphasignal.ai/news/daily-s-phonellm-matches-gpt-quality-at-94-lower-cost-for-voice-agents.md) (2026-08-24)
- [Qwen Drops Qwen3.8-Flash-Next, a 125B Open Model That Activates Only 6B Parameters](https://alphasignal.ai/news/qwen-drops-qwen3-8-flash-next-a-125b-open-model-that-activates-only-6b.md) (2026-08-24)
- [SparkLLM's Spark-X2.5 Packs 1M-Token Context Into a Tiny 4B Model](https://alphasignal.ai/news/sparkllm-s-spark-x2-5-packs-1m-token-context-into-a-tiny-4b-model.md) (2026-08-24)
- [Liquid AI's DSpark Triples LFM2.5 Speed Without Changing Its Outputs](https://alphasignal.ai/news/liquid-ai-s-dspark-triples-lfm2-5-speed-without-changing-its-outputs.md) (2026-08-20)
- [Kakao Trains a 155B MoE on 10T Tokens Without a Single Hyperparameter Sweep](https://alphasignal.ai/news/kakao-trains-a-155b-moe-on-10t-tokens-without-a-single-hyperparameter-sweep.md) (2026-08-20)
- [NVIDIA Shrinks DeepSeek V4 Flash to Run on a Single Blackwell Node](https://alphasignal.ai/news/nvidia-shrinks-deepseek-v4-flash-to-run-on-a-single-blackwell-node.md) (2026-08-19)
- [Z.ai's GLM-5.3 Tops Open-Weights Leaderboard Using Only Post-Training](https://alphasignal.ai/news/z-ai-s-glm-5-3-tops-open-weights-leaderboard-using-only-post-training.md) (2026-08-18)
- [ReliquaryForge's qwen3-4b-base-dapo-v4 Brings 32B Reasoning Power to a Single GPU](https://alphasignal.ai/news/reliquaryforge-s-qwen3-4b-base-dapo-v4-brings-32b-reasoning-power-to-a-single.md) (2026-08-18)
- [Ornith-1.5 Beats Claude Opus on Coding Benchmarks and Runs Locally](https://alphasignal.ai/news/ornith-1-5-beats-claude-opus-on-coding-benchmarks-and-runs-locally.md) (2026-08-18)
- [Solo Researcher Turns Qwen3.8 Into a Sparse MoE Running on One GPU](https://alphasignal.ai/news/solo-researcher-turns-qwen3-8-into-a-sparse-moe-running-on-one-gpu.md) (2026-08-18)
- [Ant Group's Ling 3.0 Tiny Matches GPT-120B Intelligence With 15x Fewer Parameters](https://alphasignal.ai/news/ant-group-s-ling-3-0-tiny-matches-gpt-120b-intelligence-with-15x-fewer.md) (2026-08-12)
- [NVIDIA's Nemotron 3.5 Lightning Cuts Agent Costs 58% Running 4x Faster](https://alphasignal.ai/news/nvidia-s-nemotron-3-5-lightning-cuts-agent-costs-58-running-4x-faster.md) (2026-08-11)
- [Ant Group's Ling-3.0-tiny Runs Reasoning on a MacBook at 90 tok/s](https://alphasignal.ai/news/ant-group-s-ling-3-0-tiny-runs-reasoning-on-a-macbook-at-90-tok-s.md) (2026-08-10)
- [Ant Group's Ling 3.0 Flash Beats a 1T Model With 5B Active Parameters](https://alphasignal.ai/news/ant-group-s-ling-3-0-flash-beats-a-1t-model-with-5b-active-parameters.md) (2026-08-07)