Open-Weight LLM Models — Live Ranking, Benchmarks & Specs
Explore, filter and compare open-weight language models with live specs, benchmarks and license data
Ripple-65B Preview
Ricursive Intelligence
Dense Transformer decoder with hybrid sliding-window/global attention (GQA); Gemma 3N-like, thin-and-deep design (d_model 5120, 80 layers) with FFN activation clipping, gated attention, global NoPE, and value residuals
Ripple
Apache License 2.0
Openness Score
1.0
Hydrogen-1.0
deep-principle
MoE with IndexShare sparse attention (DSA) + MTP layer
GLM
MIT License
Openness Score
1.0
SenseNova-Loop-9B-Preview
sensenova
Looped autoregressive multimodal decoder with shared looped block and two-stream manifold-constrained hyper-connections (mHC); hybrid linear/full attention; custom SenseNova Loop implementation, Qwen3.5-style text and vision configs
sensenova_loop
Apache License 2.0
Openness Score
1.0
Mistral Large 4
Mistral AI
granular MoE decoder-only transformer with MLA sparse indexed attention + Pixtral vision encoder
Mistral Large
Mistral Ecosystem Open-Weight License
Openness Score
0.7
LFM3-VL-2B
LiquidAI
LFM3 tri-hybrid LM with in-house LFM3 vision encoder and projector (vision-language)
LFM3
LFM Open License v1.0
Openness Score
0.7
LFM3-Audio-2B
LiquidAI
LFM3 backbone with causal audio encoder and LiquidCodec audio tokenizer
LFM3
LFM Open License v1.0
Openness Score
0.7
LFM3-1.7B
LiquidAI
Tri-hybrid decoder: 24 layers combining 13 gated short convolutions, 5 sliding-window attention, and 6 global attention
LFM3
LFM Open License v1.0
Openness Score
0.7
LFM3-Omni-2B
LiquidAI
LFM3 backbone with LFM3 vision encoder, LFM3 causal audio encoder, and LiquidCodec audio tokenizer
LFM3
LFM Open License v1.0
Openness Score
0.7
Nanbeige4.3-33B
Nanbeige
dense Llama-style Transformer decoder with GQA
Nanbeige4.3
Apache License 2.0
Openness Score
1.0
d1-2B
LiquidAI
LFM3 tri-hybrid vision-language backbone with D1 decision head (13 gated short convolutions, 5 sliding-window attention, 6 global attention layers)
LFM3
LFM Open License v1.0
Openness Score
0.7
Kimi K3.1 Pro
Moonshot AI
Mixture-of-Experts (MoE)
Kimi
Kimi K3.1 Pro License
Openness Score
0.7
ERNIE-5.5-Flash-Base
Baidu
MoE
ERNIE
ERNIE 5.5 Flash Base Model License Agreement v1.0
Openness Score
0.7
Spark-X2.6-4B
SparkLLM (XHToken)
Hybrid attention transformer: 1 full-attention layer combined with 3 sliding-window attention (SWA) layers, natively supporting up to 1M-token context
Spark-X2.6
Apache License 2.0
Openness Score
1.0
Flint-40B-A2.5B
Recursive
Decoder-only mixture-of-experts Transformer (hybrid sliding-window + full attention)
Flint
Apache License 2.0
Openness Score
1.0
Raon-1 mini
KRAFTON AI
hybrid Mixture-of-Experts (MoE) Transformer decoder with Raon Delta Attention (18 linear attention layers) and Gated MLA (7 full-attention layers), Raon-mHC residual connections, NoPE
Raon
Apache License 2.0
Openness Score
1.0
MiniMax-M3.1-Flash
MiniMaxAI
Native multimodal mixture-of-experts (MoE) Transformer with grouped-query attention, MiniMax Sparse Attention (MSA) throughout, and a DSpark speculative-decoding draft head
MiniMax
MiniMax Community License
Openness Score
0.7
Muse-Spark-Open-816B-A42B
Meta Superintelligence Lab
Mixture-of-Experts Transformer with Perception Encoder
Muse Spark
Muse Model License
Openness Score
0.7
Open Weights
Mamba2-Transformer Hybrid Latent Mixture of Experts (LatentMoE) with C-RADIO v4-H vision encoder
Nemotron
OpenMDW License Agreement v1.1
Openness Score
0.7