selfhost.directory

The index of self-hostable projects — ranked, categorized and served with live updates.
Own your data and stop giving it to big tech.

Alternatives to RWKV7-G1k-13.3B-20260930

120 open-weight models that do the same job as RWKV7-G1k-13.3B-20260930 (13.3B chat), ranked by how close they are in task and size, then by popularity and how recent they are. Licence and VRAM are shown for each.

the original
RWKV logo≈16 GB

This is an official BlinkDL release of RWKV-7 Goose in Hugging Face Transformers format. RWKV-7 is an attention-free recurrent architecture with a constant-size recurrent state and constant inference work per generated token. Training remains parallelizable.

13.3BApache 2.0
438Sept 2026
--vram
any
--gpus
--quant
  1. poolside logo

    Laguna-XS-2.1

    poolside·Chat

    ≈22 GB

    Laguna XS 2.1 is a 33B total parameter Mixture-of-Experts model with 3B activated parameters per token designed for agentic coding and long-horizon work on a local machine.

    33.4BA2.2B256K ctx Custom licence
    587.1KJun 2026ollama
    • Same category
    • Larger (33.4B vs 13.3B)
    • Commercial use OK
    vs
  2. ibm-granite logo

    granite-4.1-8b-base

    ibm-granite·Chat

    ≈6.7 GB

    Model Summary: Granite‑4.1‑8B‑Base is a decoder‑only language model with long‑context capabilities, designed to support a broad range of text‑to‑text generation tasks.

    8.4B128K ctx Apache 2.0
    14.4KApr 2026gguf
    • Same category
    • Similar size (8.4B vs 13.3B)
    • Commercial use OK
    vs
  3. allenai logo

    Olmo-3-7B-Instruct

    allenai·Chat

    ≈8.8 GB

    We introduce Olmo 3, a new family of 7B and 32B models both Instruct and Think variants. Long chain-of-thought thinking improves reasoning tasks like math and coding.

    7.3B64K ctx Apache 2.0
    128.7KNov 2025ollama
    • Same category
    • Similar size (7.3B vs 13.3B)
    • Commercial use OK
    vs
  4. Nemotron-3-Nano-30B-A3B-BF16 is a large language model (LLM) trained from scratch by NVIDIA, and designed as a unified model for both reasoning and non-reasoning tasks. It responds to user queries and tasks by first generating a reasoning trace and then concluding with a final response.

    31.6BA3B256K ctx nvidia-nemotron-open-model-license
    1.4MDec 2025gguf
    • Same category
    • Larger (31.6B vs 13.3B)
    • Commercial use OK
    vs
  5. poolside logo

    Laguna-XS.2

    poolside·Chat

    ≈22 GB

    Laguna XS.2 is a 33B total parameter Mixture-of-Experts model with 3B activated parameters per token designed for agentic coding and long-horizon work on a local machine. It uses Sliding Window Attention with per-head gating in 30 out of 40 layers for fast inference and low KV cache requirements.

    33.4BA2.2B256K ctx Apache 2.0
    731KApr 2026ollama
    • Same category
    • Larger (33.4B vs 13.3B)
    • Commercial use OK
    vs
  6. ByteDance-Seed logo

    Cola2

    ByteDance-Seed·Chat

    ≈6.8 GB

    Cola2 is a continuous-latent language model that co-adapts semantic compression and discrete text generation. It combines a dynamic-compression VAE, a block-causal Diffusion Transformer (DiT), and a span-causal token decoder with latent feedback.

    8.8B32K ctx Unknown licence
    1.4KNov 2026
    • Same category
    • Similar size (8.8B vs 13.3B)
    • Newer release
    vs
  7. empero-ai logo

    Qwythos-9B-v2

    empero-ai·Chat

    ≈7.2 GB

    The next iteration of Qwythos: all the reasoning of Qwythos-9B, with the looping behavior fixed. v2 keeps the deep chain-of-thought, the uncensored research posture, and the 1M-token context of its predecessor, and cleans up the rough edges that showed up in real use.

    9.7B1M ctx Apache 2.0
    1.3KJul 2026gguf
    • Same category
    • Similar size (9.7B vs 13.3B)
    • Commercial use OK
    vs
  8. ibm-granite logo

    granite-guardian-4.1-8b

    ibm-granite·Chat

    ≈6.7 GB

    Granite Guardian 4.1 8B introduces improved Bring Your Own Criteria (BYOC) support, enabling users to define arbitrary judging criteria beyond the pre-baked safety and hallucination detectors.

    8.4B128K ctx Apache 2.0
    9.2KApr 2026gguf
    • Same category
    • Similar size (8.4B vs 13.3B)
    • Commercial use OK
    vs
  9. meta-llama logo

    Llama-2-13b-chat-hf

    meta-llama·Chat

    ≈11 GB

    Llama 2 is a collection of pretrained and fine-tuned generative text models ranging in scale from 7 billion to 70 billion parameters. This is the repository for the 13B fine-tuned model, optimized for dialogue use cases and converted for the Hugging Face Transformers format.

    13B4K ctx Llama 2 Community License
    140.1KJul 2023ollama
    • Same category
    • Similar size (13B vs 13.3B)
    vs
  10. inclusionAI logo

    Ling-mini-2.0

    inclusionAI·Chat

    ≈10 GB

    Today, we are excited to announce the open-sourcing of Ling 2.0 — a family of MoE-based large language models that combine SOTA performance with high efficiency. The first released version, Ling-mini-2.0, is compact yet powerful.

    16.3BA1.4B32K ctx MIT
    4.3KSept 2025gguf
    • Same category
    • Similar size (16.3B vs 13.3B)
    • Commercial use OK
    vs
  11. deepseek-ai logo

    DeepSeek-V2-Lite

    deepseek-ai·Chat

    ≈12 GB

    Last week, the release and buzz around DeepSeek-V2 have ignited widespread interest in MLA (Multi-head Latent Attention)! Many in the community suggested open-sourcing a smaller MoE model for in-depth research. And now DeepSeek-V2-Lite comes out:

    15.7BA2.7B160K ctx DeepSeek License
    115.9KMay 2024gguf
    • Same category
    • Similar size (15.7B vs 13.3B)
    vs
  12. llm-jp logo≈12 GB

    LLM-jp-3.1 is a series of large language models developed by the Research and Development Center for Large Language Models at the National Institute of Informatics.

    13.7B4K ctx Apache 2.0
    2KMay 2025gguf
    • Same category
    • Similar size (13.7B vs 13.3B)
    • Commercial use OK
    vs
  13. utter-project logo

    EuroLLM-9B-Instruct-2512

    utter-project·Chat

    ≈7.2 GB

    This is the model card for EuroLLM-9B-Instruct-2512, an improved version of utter-project/EuroLLM-9B-Instruct. In comparison with the previous version, this version includes the long-context extension phase and the revamped post-training recipe from utter-project/EuroLLM-22B-Instruct.

    9.2B32K ctx Apache 2.0
    6.4KJan 2026gguf
    • Same category
    • Similar size (9.2B vs 13.3B)
    • Commercial use OK
    vs
  14. moonshotai logo

    Moonlight-16B-A3B

    moonshotai·Chat

    ≈12 GB

    Recently, the Muon optimizer has demonstrated strong results in training small-scale language models, but the scalability to larger models has not been proven. We identify two crucial techniques for scaling up Muon:

    16BA3B8K ctx MIT
    14KFeb 2025
    • Same category
    • Similar size (16B vs 13.3B)
    • Commercial use OK
    vs
  15. openbmb logo

    MiniCPM-SALA

    openbmb·Chat

    ≈6.4 GB

    MiniCPM-SALA (Sparse Attention and Linear Attention) is the first large-scale hybrid model effectively integrating sparse and linear attention for million-token context modeling

    9.5B512K ctx Apache 2.0
    4KFeb 2026
    • Same category
    • Similar size (9.5B vs 13.3B)
    • Commercial use OK
    vs
  16. TokenRhythm logo

    NeoHorse-1-9B

    TokenRhythm·Chat

    ≈6.8 GB

    NeoHorse-1-9B is a 9B causal language model and an initial prototype on the path toward recursive self-improvement (RSI). It is post-trained from Qwen3.5-9B for text-based agent harnesses, tool use, coding, and instruction following.

    9B256K ctx Apache 2.0
    1.4KSept 2026gguf
    • Same category
    • Similar size (9B vs 13.3B)
    • Commercial use OK
    vs
  17. deepgrove logo

    maple-preview

    deepgrove·Chat

    ≈13 GB

    Today we introduce Maple-Preview, an open-source 20B-A1B ternary-weight reasoning LLM. Maple-Preview has SOTA reasoning for its weight class and is even competitive with larger models.

    20.2BA1.5B128K ctx MIT
    2KAug 2026gguf
    • Same category
    • Similar size (20.2B vs 13.3B)
    • Commercial use OK
    vs
  18. amd logo≈12 GB

    Instella-MoE is a state-of-the-art fully open Mixture-of-Experts (MoE) language model with 16 billion total parameters and 2.8 billion active parameters trained end-to-end from pre-training to RL.

    15.9BA3B32K ctx researchrail
    221Jul 2026gguf
    • Same category
    • Similar size (15.9B vs 13.3B)
    vs
  19. ornith-ai logo

    Ornith-1.5-35B-A3B

    ornith-ai·Chat

    ≈22 GB

    Chirp Chirp! 🐦 We are introducing Ornith-1.5, a major step toward building foundation models through end-to-end self-improvement.

    36BA3B256K ctx MIT
    318.5KAug 2026ollama
    • Same category
    • Larger (36B vs 13.3B)
    • Commercial use OK
    vs
  20. allenai logo

    Olmo-3-1025-7B

    allenai·Chat

    ≈8.8 GB

    We introduce Olmo 3, a new family of 7B and 32B models. This suite includes Base, Instruct, and Think variants. The Base models were trained using a staged training approach.

    7.3B64K ctx Apache 2.0
    106.3KSept 2025gguf
    • Same category
    • Similar size (7.3B vs 13.3B)
    • Commercial use OK
    vs
  21. ServiceNow-AI logo

    Apriel-Nemotron-15b-Thinker

    ServiceNow-AI·Chat

    ≈11 GB

    Apriel-Nemotron-15b-Thinker is a 15 billion‑parameter reasoning model in ServiceNow’s Apriel SLM series which achieves competitive performance against similarly sized state-of-the-art models like o1‑mini, QWQ‑32b, and EXAONE‑Deep‑32b, all while maintaining only half the memory footprint of those…

    15B64K ctx MIT
    3.6KMay 2025gguf
    • Same category
    • Similar size (15B vs 13.3B)
    • Commercial use OK
    vs
  22. pfnet logo≈6.1 GB

    PLaMo-3-8B-Fin-Instruct is a Japanese language model for the financial domain, developed by Preferred Networks, Inc. It is fine-tuned from a PLaMo-3 base model through continued pre-training on financial corpora followed by supervised fine-tuning and direct preference optimization.

    8.5B256K ctx plamo-community-license
    983Nov 2026gguf
    • Same category
    • Similar size (8.5B vs 13.3B)
    • Newer release
    vs
  23. cais logo≈9.8 GB

    This is the official classifier for text behaviors in HarmBench. This model support standard (text) behaviors and contextual behaviors.

    13B2K ctx MIT
    41.1KFeb 2024gguf
    • Same category
    • Similar size (13B vs 13.3B)
    • Commercial use OK
    vs
  24. google logo

    shieldgemma-9b

    google·Chat

    ≈8.6 GB

    ShieldGemma is a series of safety content moderation models built upon Gemma 2 that target four harm categories (sexually explicit, dangerous content, hate, and harassment).

    9.2B8K ctx Gemma Terms of Use
    257.4KJul 2024ollama
    • Same category
    • Similar size (9.2B vs 13.3B)
    vs