We are releasing MiniCPM5-1B, the first model in the MiniCPM5 series. It is a dense 1B Transformer built for on-device, local deployment, and resource-constrained scenarios, reaching 1B-class open-source SOTA.
Crowd-v1
Vxtzq·Base model·released 21 Aug 2026
hf download Vxtzq/Crowd-v1Vxtzq's Crowd-v1 is a 1B-parameter open-weight model for chat and text generation. It takes text input and produces text output. The base variant is released under the MIT licence, which allows commercial use, and the download is not gated. The model is listed with a minimum of 1.5 GB of VRAM. It has no other sizes in its family on this page.
How much VRAM does Crowd-v1 need?
At Q4_K_M Crowd-v1 needs about 1.4 GB with an 8K context (1.4 GB at Q4, 2.8 GB at full precision). Pick your GPU to see what fits and how fast it runs.
Needs: ≈1.4 GB · Q4_K_M · 8K context — Pick a GPU to check the fit.
Advanced: context 8K · KV cache 16-bit
Smallest GPU for each quant and context (KV cache 16-bit)
| Quantisation | 4K | 8K | 16K | 32K | 64K | 128K |
|---|---|---|---|---|---|---|
| Q3_K_M | ||||||
| Q4_K_M | ||||||
| Q5_K_M | ||||||
| Q6_K | ||||||
| Q8_0 | ||||||
| FP16 |
≤16 GB card24–32 GB48–96 GBmore
| Quantisation | Total |
|---|---|
| Q2_KHeavy lossSmallest, noticeably worse quality | ≈1.2 GB |
| Q3_K_MNoticeable lossSmall, some quality loss | ≈1.3 GB |
| Q4_K_MrecommendedSmall lossRecommended default for local use | ≈1.4 GB |
| Q5_K_MNear-losslessNear-lossless for most tasks | ≈1.6 GB |
| Q6_KNear-losslessPractically lossless | ≈1.7 GB |
| Q8_0LosslessLossless in practice, twice the size of Q4 | ≈1.9 GB |
| FP16 / BF16LosslessFull precision, as released | ≈2.8 GB |
Estimates for 8K tokens of context: weights + KV cache + runtime overhead. Speed assumes memory-bound decoding at your GPU’s bandwidth. Pick a GPU above and the fit column checks every row against it.
How to pick a quant
Take the biggest quant that fits with room left for context. Q4_K_M is the usual sweet spot. Below Q4 quality drops fast, above Q6 you mostly pay memory for nothing. Big models (70B+, large MoE) tolerate low quants better than small ones.
- LosslessF16, Q8_0
- Near-losslessQ6_K, Q5_K_M
- Small lossQ5_K_S, Q4_K_M, IQ4_NL, IQ4_XS, Q4_K_S
- Noticeable lossQ4_0, Q3_K_L, Q3_K_M, IQ3_M
- Heavy lossIQ3_XXS, Q3_K_S, IQ2_M, Q2_K
Typical quality loss, not measured per model. Tiers come from KL divergence against FP16 on Llama 3 8B: llama.cpp perplexity README: LLaMA 3 8B scoreboard (KLD and ΔPPL vs FP16, Wikitext-2); llama.cpp quantize tool: ΔPPL per quant type @ Llama-3-8B
How much context fits?
At 8K context, Q4_K_M needs ≈1.4 GB(0.6 GB weights + 0.4 GB KV cache + 0.5 GB overhead). Pick a GPU above to see how much context fits.
Weights + overheadKV cache
Can the RTX 4060 run Crowd-v1? →
Speed vs quality per quant
Smaller quants run faster and fit in less memory, but lose quality.
Fits in VRAMTight fitCPU offload✕ won't runQuality is typical KL divergence vs FP16, not measured for this model.
Real-world speed of Crowd-v1
Tokens per second measured by people running it at home. A GPU shows a median once 3 reports agree on the same setup. Reports far from our estimate wait for a review. Measure yours with llama-bench →
No medians yet. Be one of the first 3 to report a setup.
All reports
Run Crowd-v1 locally
No runtime support recorded yet. Most models also run through Hugging Face Transformers.
About Crowd-v1
Crowd-v1 is the first official model architecture released for CrowdGPT, a community-driven distributed AI project. Unlike a conventional pretrained model release, Crowd-v1 is distributed with partially trained weights. The purpose of this release is to provide a common model definition and weight format that CrowdGPT clients can download and collectively train. The model is designed to be consumed by the CrowdGPT distributed training infrastructure, where individual participants contribute compute toward training a shared model. Important: Crowd-v1 undergoing pretraining. Its initial weights contain partial useful language knowledge and the model should not be expected to generate very coherent text. Crowd-v1 contains approximately 1 billion parameters. Grouped-Query Attention (GQA) Crowd-v1 uses 16 query heads but only 4 key/value heads, reducing KV-cache and attention-related memory requirements. Rotary Position Embeddings (RoPE) Rotary embeddings are applied to query and key representations. SwiGLU The feed-forward network uses a SwiGLU-style gated MLP with a hidden dimension of 2,560. Weight tying The token embedding matrix and language-model head share weights, reducing the total parameter count. Causal attention Each token can only attend to preceding tokens, making the architecture suitable for autoregressive language modeling. The released weights are generated deterministically using seed 42. Non-normalization parameters are initialized from: N(0, 0.02)
Full model card on Hugging Face →Alternatives to Crowd-v1
All alternatives →Qwen3 is the latest generation of large language models in Qwen series, offering a comprehensive suite of dense and mixture-of-experts (MoE) models.
K2-Horizon-0.9B is the compact dense member of the K2-Horizon family: a 0.9B-class decoder-only model with a 128K context window.
LFM2.5 is a new family of hybrid models designed for on-device deployment. It builds on the LFM2 architecture with extended pre-training and reinforcement learning.
Gemma is a family of lightweight, state-of-the-art open models from Google, built from the same research and technology used to create the Gemini models.
Limite 1B - Violetto is Paradigma’s first model, designed for high-throughput solutions of difficult mathematical problems.
Crowd-v1 FAQ
- How much VRAM does Crowd-v1 need?
- About 1.4 GB at Q4_K_M with an 8K context. Q8_0 needs about 1.9 GB and full precision about 2.8 GB. Longer contexts need more memory for the KV cache.
- Can Crowd-v1 run on the RTX 4060?
- Yes. It fits at FP16 / BF16 (about 2.8 GB), at an estimated 75 tokens/s. It fits comfortably on the Apple M4 (16 GB), Radeon RX 6700 XT, RTX 3060 12GB.
- Can I use Crowd-v1 commercially?
- Yes. It is released under MIT. Permissive: do almost anything as long as you keep the copyright notice.
Rate this model
Data from Hugging Face, the model card and public leaderboards. Memory and speed figures are estimates; real numbers depend on the runtime, context length and drivers. Always check the licence yourself before commercial use.
0 comments