Language Models ✕

Language Models

underdogs breaking out
01
UncensoredLabs/uncensored-ai-api
+97% /wk +120 ★/day→steady

UncensoredLabs packages refusal-stripped DeepSeek and GLM models behind a drop-in base-URL swap — the repo itself is docs and sample code, not the service.

★ 867 JavaScript Inference · Serving · explained
02
qingjian-team/qingjian
+81% /wk +889 ★/day↗accelerating

A Rust-based Chinese IME that treats your daily typing as a language lesson, because opening a study app is a chore.

★ 7.7k Rust Language Models · explained Feature
03
mw00/project-maya
+65% /wk +250 ★/day↗accelerating

Project Maya runs GLM-5.3-Flash privately on one to sixteen GPUs by keeping hot experts in VRAM, warm ones in RAM, and the rest on NVMe.

★ 2.7k Cuda Inference · Serving · explained
04
nokia-applied-research/AnyJev
+47% /wk +108 ★/day↗accelerating

AnyJev turns any open LLM into a calibrated decision engine—typed choices with probabilities you can threshold—without touching a single model weight.

★ 1.6k Python LLMOps · Eval · explained
05
youngyangyang04/llm-master
+43% /wk +96 ★/day↗accelerating

It turns 150+ scattered LLM guides into a single engineering curriculum—from API basics to production Agents and ByteDance interviews—because hoarding bookmarks is not a learning strategy.

★ 1.6k Learning · explained
06
divagr18/qwenputer
+34% /wk +37 ★/day→steady

An experimental Qwen3.5-4B language model with a 32-bit RISC-V CPU compiled directly into its transformer weights, using the model's Gated DeltaNet state as RAM.

★ 756 Python Language Models · explained
07
MemTensor/MemT-1.0
+33% /wk +25 ★/day→steady

MemT-1.0 bakes persistent memory into the model itself, so history lives in a fixed-size latent state instead of a growing KV cache.

★ 528 Python Language Models · explained
09
wfzyx/von
+30% /wk +40 ★/day↗accelerating

It replaces autoregressive LLM classification with a single parallel forward pass, cutting latency from seconds to sub-20 ms while staying local.

★ 919 Python Language Models · explained
10
peonist-ai/halogen-flash-server
+30% /wk +53 ★/day↗accelerating

It rewrites every kernel for AMD Strix Halo and a single 179B-parameter MoE family to beat general-purpose engines without spending fewer bits.

★ 1.3k Shell Inference · Serving · explained
11
hello-diana/MASCOT
+23% /wk +26 ★/day↗accelerating

MASCOT is a training framework that keeps multi-agent LLM companions in distinct personas while a director agent prevents them from collapsing into an echo chamber.

★ 804 Python Agents · explained
12
FireRedTeam/FireRedASR2S
+21% /wk +22 ★/day↗accelerating

FireRedASR2S bundles ASR, voice detection, language ID, and punctuation into a single pipeline optimized for Mandarin, 20+ regional variants, English code-switching, and even singing.

★ 733 Python Image · Video · Audio · explained
13
techjarves/Portable-Local-Studio
+20% /wk +57 ★/day↗accelerating

It unifies Stable Diffusion, GGUF chat, Whisper, and Kokoro TTS into a single offline desktop GUI so you can skip cloud APIs, subscriptions, and censorship filters.

★ 2k JavaScript Inference · Serving · explained
14
warpfront/hipfire
+20% /wk +23 ★/day↗accelerating

Most LLM inference stacks treat consumer AMD GPUs as second-class citizens; hipfire targets the full RDNA family with a single Rust binary that keeps Python out of the hot path.

★ 806 Rust Inference · Serving · explained
15
rednote-machine-learning/RedKnot
+18% /wk +79 ★/day↗accelerating

RedKnot accelerates long-context inference by sorting attention heads into four species—global, local, retrieval, and dense—then giving each its own KV reuse strategy and sparsity rules, built as an SGLang extension.

★ 3.1k Python Inference · Serving · explained
16
StartLuxLabs/StartLux-Decision
+17% /wk +42 ★/day→steady

StartLux-Decision treats decisions as classification, not chat: one forward pass, a calibrated probability per option, and reasoning only when confidence dips below a gate.

★ 1.7k Python Language Models · explained
17
syv-ai/HyperQwen
+17% /wk +63 ★/day↗accelerating

This repo is a surgical stack of vLLM patches, requantization scripts, and speculative decoders built to squeeze Qwen3.8-27B — and up to 268k tokens of context — into a single 24 GB consumer GPU.

★ 2.6k Python Inference · Serving · explained
18
receptron/laya
+14% /wk +20 ★/day→steady

Laya runs in Node.js via ONNX Runtime, turning tickets and emails into calibrated choices, rubric scores, and P(true) values in a single forward pass.

★ 990 TypeScript Inference · Serving · explained
19
cobanov/awesome-jev
+14% /wk +11 ★/day→steady

A curated, source-backed catalog of 155 projects built on Jev, a model that answers typed questions with probabilities instead of generating text.

★ 588 Learning · explained
20
OmniJev/awesome-jev-gallery
+13% /wk +9.8 ★/day→steady

A curated awesome-list tracking 167 papers, models, benchmarks, and independent evaluations around TypeSafe's System One decision models.

★ 510 Python Learning · explained
loading more…

heatdrop uses Google Analytics to see which pages get read — nothing else. Your call. How we handle data.