moe · 115 topics

deploying qwen3 8 2 4t a95b on amazon sagemaker hyperpod with vllm

k2 horizon mova 36b a4b

nvidia nemotron ioi 2026

mbzuai k2 horizon 375b

pipecat ai phonellm alpha 1

z ai glm 5 3 open weights

tencent hy4 preview

glm 5 3 flash

glm 5 3 flash performance analysis

qwen38 flash next

z ai glm 5 3 flash matches claude opus

30b open models trend

ornith 1 5 35b a3b gguf

ornith 1 5 35b a3b

glm 5 3

qwen3 8 2 4t a95b fp8

nvidia nemotron 3 5 lightning 30b a3b nvfp4

inclusionai ling 3 0 tiny model

glm 5 3 cybersecurity ai

alibaba opens qwen3 8 max weights

deepseek v4 pro release

qwen3 8 2 4t a95b

nvidia nemotron 3 5 lightning

deepseek v4 flash debugging benchmark

ling 3 0 flash

bytedance trains massive ai model

inclusionai ling 3 0 flash

latest open artifacts 23

thinkingmachines inkling small

kimi k3

luffythefox qwen36 35b genesis

deepseek v4 flash agent benchmarks

thinking machines inkling small

tencent open sources angelspec

moonshot ai kimi k3 2 8t model

motif 3 beta release

kwaipilot kat coder v2 5 dev

poolside laguna s 2.1

upstage solar open2 250b

kimi k3 28t a50b

moonshot ai kimi k3

mira muratis thinking machines drops inkling

why performance per watt is the ultimate metric for ai infrastructure efficiency

nvidia nemotron labs audex 30b a3b

unsloth deepseek v4 flash gguf

hf trending deepseek v4 pro dspark

gpt 5 6 sol deepseek dspark

glm 5 2 beats claude

accelerating transformers fine tuning with nvidia nemo automodel

deepseek v4 efficient million token

glm 5 2 open weights model

nvidia blackwell mlperf training 6 0

minimax m3

google diffusiongemma high speed inference

diffusiongemma 4x faster text generation

cohere north mini code

2026 06 08 papers 2412.19437

microsoft mai thinking 1 moe

jetbrains mellum2 12b moe

hauhaucs qwen3 6 35b a3b uncensored

laguna m1 xs2

cohere command a plus 218b moe

allen institute emo moe modularity

ai2 emo moe

inclusionai ling 2 6 flash release

nvidia nemotron 3 nano omni

deepseek v4 flash

nvidia nemotron 3 nano omni 30b a3b reasoning bf16

deepseek v4 series release

xiaomimimo mimo v2 5

xiaomimimo mimo v2.5 pro

nvidia nemotron 3 nano omni

nvidia nemotron 3 nano omni multimodal efficiency

deepseek v4 pro preview

google gemma 4 31b it trending

deepseek v4 agentic long context

takeaways discussion about the deepseek v4 architecture

smelt scaling laws for compute matched moe looped transformers

jina ocr v1

best practice critic optimization

constant competitive algorithm for dynamic mixture of experts serving

depth aware sensitivity analysis of moe models

dual flow transformers

towards physics of multimodal pretraining

arex recursively self improving agent deep research

slai t rex deepseek v4 ascend superpod

is moe routing a huffman code

solar open 2 technical report

gemma 4 technical report

path constrained mixture of experts

deepseek v4 technical report fp4 qat

deltafin kimi k3 m5 max

coheres open source megakernel beats vllm

benchmarking small llm inference on sagemaker ai g7 vs g5 and g6

slotstream qwen3 8 flash next

perplexity lily qwen36

vllm v0 28 0 sparse attention speculative decoding

prime flash moe

unsloth kimi k3 gguf

waste kimi k3 engine

poolside laguna s 2 1 rate limits

deploying kimi k3 on aws

moonshotai kimi k3

poolside laguna s 21 nvfp4

unsloth laguna s 2 1 gguf

laguna s 2 1

running gemma 4 26b on xeon

colibri glm 5 2 runner

tencent hy3

qwen agentworld 35b a3b

internscience agents a1

liquid ai lfm2 5 8b a1b 38t moe

qwen3 7 max agent frontier

microsoft github copilot polaris

thinking machines inkling release