Goblin
News
AI news, distilled.
|
News
About
News
About
Filtered by:
inference-efficiency
Clear
Titles
Summaries
8
NVIDIA Releases Nemotron-3.5-Super-VL-120B-A12B Multimodal Model Family
Models
2
6h ago
8
NVIDIA Releases Nemotron-3.5-Super-VL-120B-A12B Multimodal Model Family
Models
· 2 srcs · 6h ago
Discuss
8
StepFun: Step 5 Preview, a 600B Sparse MoE Model Built for Agentic Work
Models
2
14h ago
8
StepFun: Step 5 Preview, a 600B Sparse MoE Model Built for Agentic Work
Models
· 2 srcs · 14h ago
Discuss
7
NVIDIA: Nemotron 3.5 Super VL Open 120B Agent Model Launches With Two Edge 'Bolt' Models
Models
2
5h ago
7
NVIDIA: Nemotron 3.5 Super VL Open 120B Agent Model Launches With Two Edge 'Bolt' Models
Models
· 2 srcs · 5h ago
Discuss
7
vLLM: Bitwise-Consistent RL Training Removes Train-Inference Drift, Holds Reward Where Baseline Collapses
Products
1
13h ago
7
vLLM: Bitwise-Consistent RL Training Removes Train-Inference Drift, Holds Reward Where Baseline Collapses
Products
· 1 src · 13h ago
Discuss
7
IBM: Granite Switch 4.1 Combines Chat, Decision, Guardian and Speech Adapters in One Model
Models
1
6h ago
7
IBM: Granite Switch 4.1 Combines Chat, Decision, Guardian and Speech Adapters in One Model
Models
· 1 src · 6h ago
Discuss
7
Krea 3: Image Editing Model Claims Sub-8-Second Edits, 4x Faster Than Nano Banana Pro
Models
1
6h ago
7
Krea 3: Image Editing Model Claims Sub-8-Second Edits, 4x Faster Than Nano Banana Pro
Models
· 1 src · 6h ago
Discuss
7
vLLM: Day 0 Support for Meta's Muse Spark Open, an 816B MoE with 1M Context
Open Source
1
5h ago
7
vLLM: Day 0 Support for Meta's Muse Spark Open, an 816B MoE with 1M Context
Open Source
· 1 src · 5h ago
Discuss
6
llama.cpp: Merge-Ready Support for Unannounced 'maion-coder' Architecture, Tied to Microsoft MAI-Code-1.1-Flash-On-Device
Open Source
1
13h ago
6
llama.cpp: Merge-Ready Support for Unannounced 'maion-coder' Architecture, Tied to Microsoft MAI-Code-1.1-Flash-On-Device
Open Source
· 1 src · 13h ago
Discuss
6
Apple Research: Selection-Based Reasoning Cuts Small Multimodal Agent Latency Over 90% Per Turn
Research
1
5h ago
6
Apple Research: Selection-Based Reasoning Cuts Small Multimodal Agent Latency Over 90% Per Turn
Research
· 1 src · 5h ago
Discuss
8
SpaceXSI: Grok Voice 3.0 Launches at $0.08/min, Tops Artificial Analysis Speech-to-Speech Index
Products
1
1d ago
8
SpaceXSI: Grok Voice 3.0 Launches at $0.08/min, Tops Artificial Analysis Speech-to-Speech Index
Products
· 1 src · 1d ago
Discuss
8
Fish Audio: Drama 3 Open-Weight TTS Model Takes Top Open-Weight Spot on Artificial Analysis Leaderboard
Models
1
1d ago
8
Fish Audio: Drama 3 Open-Weight TTS Model Takes Top Open-Weight Spot on Artificial Analysis Leaderboard
Models
· 1 src · 1d ago
Discuss
7
pg_vectorsearch: Alpha PostgreSQL Extension for Approximate Nearest-Neighbor Search with PRISM Index
Products
1
1d ago
7
pg_vectorsearch: Alpha PostgreSQL Extension for Approximate Nearest-Neighbor Search with PRISM Index
Products
· 1 src · 1d ago
Discuss
6
Lakebase Search: Sizing Guidance from a 700-Deck Hybrid Search App and a 4M-Passage Test
Products
1
1d ago
6
Lakebase Search: Sizing Guidance from a 700-Deck Hybrid Search App and a 4M-Passage Test
Products
· 1 src · 1d ago
Discuss
6
TypeSafe: Jev 1.14.0 Released With Experimental 'Danger Zone' Features
Models
1
1d ago
6
TypeSafe: Jev 1.14.0 Released With Experimental 'Danger Zone' Features
Models
· 1 src · 1d ago
Discuss
6
vLLM: Push-Based NIXL Connector Speeds Prefill-Decode Disaggregated Serving
Infra
1
1d ago
6
vLLM: Push-Based NIXL Connector Speeds Prefill-Decode Disaggregated Serving
Infra
· 1 src · 1d ago
Discuss
6
vLLM: Adaptive Verification Now Works Without a Drafter Confidence Head
Products
1
1d ago
6
vLLM: Adaptive Verification Now Works Without a Drafter Confidence Head
Products
· 1 src · 1d ago
Discuss
6
AWS: SageMaker JumpStart Tutorial Covers Deploying and Evaluating Models with Python SDK v3
Products
1
1d ago
6
AWS: SageMaker JumpStart Tutorial Covers Deploying and Evaluating Models with Python SDK v3
Products
· 1 src · 1d ago
Discuss
7
AnythingLLM Desktop: Replaces Bundled Ollama with llama.cpp Run via AMD's Lemonade
Products
1
2d ago
7
AnythingLLM Desktop: Replaces Bundled Ollama with llama.cpp Run via AMD's Lemonade
Products
· 1 src · 2d ago
Discuss
7
Liquid AI: LFM2.5-24B-A2B Released as Largest and Final Model in LFM2.5 Family
Models
3
2d ago
7
Liquid AI: LFM2.5-24B-A2B Released as Largest and Final Model in LFM2.5 Family
Models
· 3 srcs · 2d ago
Discuss
7
Microsoft Launches Decision-1, a Low-Cost Model Returning Calibrated Probabilities for Predefined Answers
Models
4
2d ago
7
Microsoft Launches Decision-1, a Low-Cost Model Returning Calibrated Probabilities for Predefined Answers
Models
· 4 srcs · 2d ago
Discuss
7
Janas-LLM: Dependency-Free C Engine Streams MoE Experts from SSD to Run 80B Models on Laptops
Open Source
1
2d ago
7
Janas-LLM: Dependency-Free C Engine Streams MoE Experts from SSD to Run 80B Models on Laptops
Open Source
· 1 src · 2d ago
Discuss
7
SelfCompact: Language Model Agents Learn to Decide When to Compress Their Own Context
Research
1
2d ago
7
SelfCompact: Language Model Agents Learn to Decide When to Compress Their Own Context
Research
· 1 src · 2d ago
Discuss
7
GitHub: ModernBERT Secret Classifier Replaces LLM Pipeline for Generic Secrets Detection
Products
1
2d ago
7
GitHub: ModernBERT Secret Classifier Replaces LLM Pipeline for Generic Secrets Detection
Products
· 1 src · 2d ago
Discuss
7
Intel: Crescent Island Inference GPU Becomes Data Center GPU D-Series (D270), 480 GB LPDDR5X per Card
Infra
3
2d ago
7
Intel: Crescent Island Inference GPU Becomes Data Center GPU D-Series (D270), 480 GB LPDDR5X per Card
Infra
· 3 srcs · 2d ago
Discuss
7
OpenBMB: MiniCPM-V-4.7-1B, a 1B-Parameter Image-Text-to-Text Model, Released with Multi-Runtime Support
Models
1
2d ago
7
OpenBMB: MiniCPM-V-4.7-1B, a 1B-Parameter Image-Text-to-Text Model, Released with Multi-Runtime Support
Models
· 1 src · 2d ago
Discuss
7
AWS: 'Priming' Converts Pretrained Transformers into Hybrid SSM Models at Under 1% of Pretraining Cost
Research
1
2d ago
7
AWS: 'Priming' Converts Pretrained Transformers into Hybrid SSM Models at Under 1% of Pretraining Cost
Research
· 1 src · 2d ago
Discuss
6
LangChain Adds ScaleDown Constrained Decoding Models to Deep Agents and LangSmith Gateway
Products
1
2d ago
6
LangChain Adds ScaleDown Constrained Decoding Models to Deep Agents and LangSmith Gateway
Products
· 1 src · 2d ago
Discuss
6
Amazon SageMaker AI: Fused MoE Kernel Optimizations, Including Swap-AB for NVIDIA Hopper
Infra
1
2d ago
6
Amazon SageMaker AI: Fused MoE Kernel Optimizations, Including Swap-AB for NVIDIA Hopper
Infra
· 1 src · 2d ago
Discuss
6
Orca Security: Discovery AI Natural-Language Cloud Query Agent Built on Amazon Bedrock
Products
1
2d ago
6
Orca Security: Discovery AI Natural-Language Cloud Query Agent Built on Amazon Bedrock
Products
· 1 src · 2d ago
Discuss
6
Minions: Paper Proposes Distributed Virtual GPU Cloud Built on Idle Apple Silicon Devices
Research
2
2d ago
6
Minions: Paper Proposes Distributed Virtual GPU Cloud Built on Idle Apple Silicon Devices
Research
· 2 srcs · 2d ago
Discuss
6
Nvidia: DLSS 5 Tones Down Face Changes for Hogwarts Legacy; October Update Up to 2x Faster on RTX 50
Products
1
2d ago
6
Nvidia: DLSS 5 Tones Down Face Changes for Hogwarts Legacy; October Update Up to 2x Faster on RTX 50
Products
· 1 src · 2d ago
Discuss
6
Four Used Tesla V100s on NVLink Board Outrun Two DGX Sparks and an M5 Ultra on Local LLM Inference
Infra
1
2d ago
6
Four Used Tesla V100s on NVLink Board Outrun Two DGX Sparks and an M5 Ultra on Local LLM Inference
Infra
· 1 src · 2d ago
Discuss
7
Z.ai Releases GLM-Legal, a Gated Legal Model Built on GLM-5.3-Flash
Updated Oct 13
Models
3
2d ago
7
Z.ai Releases GLM-Legal, a Gated Legal Model Built on GLM-5.3-Flash
Upd Oct 13
Models
· 3 srcs · 2d ago
Discuss
7
AWS: Multi-Node Ray LLM Deep Learning Containers Arrive for Amazon EKS with Disaggregated Serving
Infra
1
3d ago
7
AWS: Multi-Node Ray LLM Deep Learning Containers Arrive for Amazon EKS with Disaggregated Serving
Infra
· 1 src · 3d ago
Discuss
7
Microsoft Announces Maia 300 Inference Chip, Plans Multi-Gigawatt Maia Rollout From 2027
Infra
7
3d ago
7
Microsoft Announces Maia 300 Inference Chip, Plans Multi-Gigawatt Maia Rollout From 2027
Infra
· 7 srcs · 3d ago
Discuss
7
vLLM: Bi-Directional KV Transfer Cuts Multi-Turn Agentic TTFT by Up to 3.3x
Products
1
3d ago
7
vLLM: Bi-Directional KV Transfer Cuts Multi-Turn Agentic TTFT by Up to 3.3x
Products
· 1 src · 3d ago
Discuss
6
AWS: llama.cpp Container on SageMaker AI Serves Quantized and Decision Models
Products
1
3d ago
6
AWS: llama.cpp Container on SageMaker AI Serves Quantized and Decision Models
Products
· 1 src · 3d ago
Discuss
6
Speculative Decoding Explained: Memory-Bound LLM Decoding and Drafters from EAGLE-3 to DFlash
Products
1
3d ago
6
Speculative Decoding Explained: Memory-Bound LLM Decoding and Drafters from EAGLE-3 to DFlash
Products
· 1 src · 3d ago
Discuss
6
Lovable: Enterprise EU Inference Setting Routes AI Requests to European Endpoints
Products
1
3d ago
6
Lovable: Enterprise EU Inference Setting Routes AI Requests to European Endpoints
Products
· 1 src · 3d ago
Discuss
6
IIT Guwahati, IIT Kharagpur: Ultrathin Light-Sensing Device Fuses Vision, Memory and Processing at 500x Lower Energy
Research
1
3d ago
6
IIT Guwahati, IIT Kharagpur: Ultrathin Light-Sensing Device Fuses Vision, Memory and Processing at 500x Lower Energy
Research
· 1 src · 3d ago
Discuss
6
llama.cpp: PR #29600 Adds Runtime Support for Prism Bonsai 2 27B Ternary Model
Open Source
1
3d ago
6
llama.cpp: PR #29600 Adds Runtime Support for Prism Bonsai 2 27B Ternary Model
Open Source
· 1 src · 3d ago
Discuss
6
Amazon Bedrock: OpenAI GPT-6 Sol, Luna and Astra Now Available via Global Cross-Region Inference
Products
1
3d ago
6
Amazon Bedrock: OpenAI GPT-6 Sol, Luna and Astra Now Available via Global Cross-Region Inference
Products
· 1 src · 3d ago
Discuss
6
AuroraIMG-6M: 6.8M-Parameter Text-to-Image Model Trained From Scratch, Runs on CPU
Models
1
3d ago
6
AuroraIMG-6M: 6.8M-Parameter Text-to-Image Model Trained From Scratch, Runs on CPU
Models
· 1 src · 3d ago
Discuss
6
Microsoft: Surface Laptop Ultra RTX Spark 128GB/1TB ($5,899.99) listed out of stock
Products
1
3d ago
6
Microsoft: Surface Laptop Ultra RTX Spark 128GB/1TB ($5,899.99) listed out of stock
Products
· 1 src · 3d ago
Discuss
6
Voice AI Execs Say Field Lacks Its 'ChatGPT Moment'; Speed and Transcription Accuracy Seen as Key Gaps
Products
1
4d ago
6
Voice AI Execs Say Field Lacks Its 'ChatGPT Moment'; Speed and Transcription Accuracy Seen as Key Gaps
Products
· 1 src · 4d ago
Discuss
7
vLLM: Day-0 Support for NVIDIA Vera Rubin NVL72, Early Results Show Up to 7.8x Per-GPU Throughput Gain
Infra
1
5d ago
7
vLLM: Day-0 Support for NVIDIA Vera Rubin NVL72, Early Results Show Up to 7.8x Per-GPU Throughput Gain
Infra
· 1 src · 5d ago
Discuss
7
Google AI Edge ML Drift: GPU inference engine for on-device generative models
Open Source
1
5d ago
7
Google AI Edge ML Drift: GPU inference engine for on-device generative models
Open Source
· 1 src · 5d ago
Discuss
6
Easy Fox: Free Steam Demo Reportedly Costs $1,000+ a Day in AI Fees, Forcing a Bank Loan
Products
1
5d ago
6
Easy Fox: Free Steam Demo Reportedly Costs $1,000+ a Day in AI Fees, Forcing a Bank Loan
Products
· 1 src · 5d ago
Discuss
6
vLLM merges /v1/systemone endpoint returning per-answer probabilities for typed questions
Open Source
1
5d ago
6
vLLM merges /v1/systemone endpoint returning per-answer probabilities for typed questions
Open Source
· 1 src · 5d ago
Discuss
8
OpenAI Rolls Out Ultrafast Mode for GPT-6.1 Sol, Up to 8x Faster
Products
1
6d ago
8
OpenAI Rolls Out Ultrafast Mode for GPT-6.1 Sol, Up to 8x Faster
Products
· 1 src · 6d ago
Discuss
Filters
Signal
Title
Category
Sources
Posted
Discuss
