Trending

Show HN: GPT-2 implemented using graphics shaders

Suno v4.5

Claude can now connect to your world

RustAssistant: Using LLMs to Fix Compilation Errors in Rust Code

LLMs for Engineering: Teaching Models to Design High Powered Rockets

Llasa: Llama-Based Speech Synthesis

XAI Dev Leaks API Key for Private SpaceX, Tesla LLMs

When ChatGPT Broke an Entire Field: An Oral History

Mercury, the first commercial-scale diffusion language model

Phi-3 is making small language models actually useful

Two Ais Talking in real time

LLM GPU calculator for inference and fine-tuning requirements

Looking for some advice on choosing between Gemini and Llama for my AI project.

[R] Reinforcement Learning for Reasoning in Large Language Models with One Training Example

Granite-4-Tiny-Preview is a 7B A1 MoE

Qwen3 0.6B running at ~75 tok/s on IPhone 15 Pro

Impressive Qwen 3 30 MoE

Microsoft just released Phi 4 Reasoning (14b)

Nari Labs

Save item

Dia-1.6B

Qwen

Save item

Qwen3-235B-A22B

Task: text-generation

Qwen

Save item

Qwen3-30B-A3B

Task: text-generation

Qwen

Save item

Qwen3-32B

Task: text-generation

TNG Technology Consulting GmbH

Save item

DeepSeek-R1T-Chimera

Task: text-generation

Qwen

Save item

Qwen3-8B

Task: text-generation

Qwen

Save item

Qwen3-0.6B

Task: text-generation

JetBrains

Save item

Mellum-4b-base

Task: text-generation

Microsoft

Save item

bitnet-b1.58-2B-4T

Task: text-generation

Giskard-AI

Save item

giskard

🐢 Open-Source Evaluation & Testing for AI & LLM systems

BlinkDL

Save item

RWKV-LM

RWKV (pronounced RwaKuv) is an RNN with great LLM performance, which can also be directly trained like a GPT transformer (parallelizable). We are at RWKV-7 "Goose". So it's combining the best of RNN and transformer - great performance, linear time, constant space (no kv-cache), fast training, infinite ctx_len, and free sentence embedding.