In-training periodic checkpoint (step 16300). Architecture: Kimi K3 mini port (KimiLinearForCausalLM). Not a finished model -- intermediate weights uploaded during training for safekeeping/diffing.
EdgeRock-4B-NVFP4 is an open-weight model for text generation from Aaron Xu, released under other. It has 2.6B parameters and a 262,144-token context. At 16-bit it needs about 6.3 GB of GPU memory, which fits on 1x MI300X from $1.85 an hour, at the lowest prices in the SAVRN Index. It draws 24 downloads a month.
EdgeRock's constant-memory reasoning architecture with packed NVFP4 weights. Fixed memory for long-running generation.
Parameters2.6B
Context262,144
Weights3.5 GB
Licenseother
AccessOpen weights
Monthly Downloads24
Runs On
What it takes to serve EdgeRock-4B-NVFP4 (2.6B parameters): the memory its weights need at each precision, and the cheapest way to rent enough data-center GPUs to hold them.
| Precision | Weights | Memory needed | Cheapest setup | Per hour | Also fits |
|---|---|---|---|---|---|
| 16-bit | 5.2 GB | 6.3 GB | 1x MI300X (192 GB) Vultr |
$1.85 | 1x H100 $1.99 · 1x MI325X $2.00 |
| 8-bit | 2.6 GB | 3.1 GB | 1x MI300X (192 GB) Vultr |
$1.85 | 1x H100 $1.99 · 1x MI325X $2.00 |
| 4-bit | 1.3 GB | 1.6 GB | 1x MI300X (192 GB) Vultr |
$1.85 | 1x H100 $1.99 · 1x MI325X $2.00 |
Memory is the weights at that precision plus 20% for the runtime and a short context; a long context needs more. Prices are the lowest on-demand hourly rates in the SAVRN Index, read Oct 23, 2026.
EdgeRock-4B-NVFP4 on every accelerator the SAVRN Index prices, at every precision
Model Card
EdgeRock's constant-memory reasoning architecture with packed NVFP4 weights. Fixed memory for long-running generation. This EdgeRock-4B deployment package retains recurrent state, local attention, and Rock Memory Anchors (RMA), keeping persistent history at a fixed capacity as context grows. 固定记忆，长程生成。 本版本保留 EdgeRock-4B 的递推状态、局部注意力和记忆岩锚（RMA），历史内存不随上下文增长。 Rock Matrix Whitepaper / 岩芯矩阵技术白皮书： whitepaper connects EdgeRock's fixed memory, DiffusionRock's parallel generation, and SiliconRock's model-native silicon into the route toward Ultra-Fast Token. 白皮书讲清固定记忆、并行生成、模型固化三条路线，以及它们如何走向更快、更经济的生成。 This text checkpoint uses NVIDIA Model Optimizer's W4A16 NVFP4 format: E2M1 4-bit weights, one FP8…
Excerpt from the card by Aaron Xu, licensed other.
Configuration
- Architecture
- EdgeRockForCausalLM
- Context length (tokens)
- 262,144
- Layers
- 32
- Hidden size
- 2,560
- Feed-forward size
- 9,216
- Attention heads
- 16
- Key/value heads
- 4
- Head dimension
- 256
- Vocabulary size
- 248,320
- Model type
- edgerock
- Quantization
- modelopt
Identity and Version
- Repository
- hkust-aaron/EdgeRock-4B-NVFP4
- Publisher
- Aaron Xu
- Task
- Text generation
- Modality
- Text
- Library
- transformers
- Parameters
- 2.6B parameters
- Languages
- Not stated by the source
- Revision
- 1fa8924e5f51830bcc1d710987b97a5f15aed0e1
- First published
- 2026-10-19
- Last updated
- 2026-10-23
Files and Weights
22 files, 3.5 GB in total. The weights are 1 file totalling 3.5 GB in safetensors.
Weights1 file · 3.5 GB
Configuration11 files · 175.5 KB
Tokenizer2 files · 20.0 MB
Documentation4 files · 38.0 KB
Other3 files · 2.1 KB
Repository1 file · 1.6 KB
Every file
| File | Type | Size | SHA-256 |
|---|---|---|---|
| model.safetensors | Weights | 3.5 GB | 889d2e7b6c8e |
| config.json | Configuration | 8.2 KB | — |
| configuration_edgerock.py | Configuration | 945 B | — |
| conversion.json | Configuration | 41.7 KB | — |
| generation_config.json | Configuration | 168 B | — |
| hf_quant_config.json | Configuration | 4.9 KB | — |
| load_nvfp4.py | Configuration | 2.1 KB | — |
| modeling_edgerock.py | Configuration | 67.6 KB | — |
| nvfp4_config.json | Configuration | 18.9 KB | — |
| provenance.json | Configuration | 2.5 KB | — |
| quantize.py | Configuration | 5.1 KB | — |
| validation.json | Configuration | 23.6 KB | — |
| LICENSE | Documentation | 18.0 KB | — |
| LICENSE-CODE | Documentation | 11.4 KB | — |
| NOTICE | Documentation | 3.6 KB | — |
| README.md | Documentation | 4.9 KB | — |
| SHA256SUMS | Other | 1.6 KB | — |
| chat_template.jinja | Other | 217 B | — |
| requirements.txt | Other | 251 B | — |
| .gitattributes | Repository | 1.6 KB | — |
| tokenizer.json | Tokenizer | 20.0 MB | 06b9509352d2 |
| tokenizer_config.json | Tokenizer | 1.1 KB | — |
License and Download
- License
- other
- Access
- Open weights, no gate
- Download size
- 3.5 GB
Download from Aaron Xu
Released by Aaron Xu through its official repository on Hugging Face.
Built From
- Derived from rockmatrix-ai/EdgeRock
- Quantized from rockmatrix-ai/EdgeRock
Memory Requirements
| Precision | Weights in memory |
|---|---|
| As published | 3.5 GB |
| 16-bit | 5.2 GB |
| 8-bit | 2.6 GB |
| 4-bit | 1.3 GB |
Weights only, from the published parameter count; the key-value cache and runtime add to this.
Questions About EdgeRock-4B-NVFP4
How much GPU memory does EdgeRock-4B-NVFP4 need?
About 6.3 GB at 16-bit and 1.6 GB at 4-bit: the weights (2.6B parameters) plus a working margin. A long context needs more.
What is the cheapest GPU to run EdgeRock-4B-NVFP4 on?
At 16-bit, 1x MI300X from $1.85 an hour; at 4-bit, 1x MI300X from $1.85 an hour, at the lowest on-demand prices the SAVRN Index lists.
What license is EdgeRock-4B-NVFP4 released under?
other, as its publisher declares it. Read the license text before commercial use.
What is EdgeRock-4B-NVFP4's context length?
262,144 tokens, from the maximum position embeddings in its published configuration.
Similar Models
An English LFM2.5 2.6B checkpoint associated with the Trendyol Cybersecurity Instruction Tuning Dataset. This repository contains both Transformers-format model files and local GGUF exports. This is a research release. The repository does not currently publish benchmark, baseline-comparison, or safety-evaluation results. Transformers assets include model.safetensors, configuration files, tokenizer files, and a chat template. Keeping both formats in one repository is convenient, but users should explicitly choose the path that matches their runtime. The checked configuration includes a bitsandbytes quantization block. Pin and test the exact Transformers, Accelerate, bitsandbytes, and device…
A JANG bundle of LiquidAI/LFM2.5-24B-A2B: Liquid AI's 24B mixture-of-experts reasoning model with 2B active parameters (40 layers: 30 short-convolution and 10 full-attention, 64 routed experts with 4 active, two dense layers), built for on-device agents. It always thinks before answering and calls tools in Liquid's Pythonic format. 44.7 GiB of bf16 weights in 7.3 GiB. Every quantized matrix with an input Hessian was rounded with GPTQ. The Hessians come from the model's own outputs: the bf16 model wrote its thinking and answers for math, finance, code in eleven languages, terminal and coding-agent sessions (with real tool results), multi-step tool calls, networking, cybersecurity, science…
English | We are releasing MiniCPM5-2B, the second model in the MiniCPM5 series, following MiniCPM5-1B. It is a dense 2B Transformer that scales up the same training recipe, built for on-device, local deployment, and resource-constrained scenarios, reaching 2B-class open-source SOTA. 2B-class open-source SOTA: compared with strong open-source models of similar size, MiniCPM5-2B achieves SOTA performance within this comparison set. It remains competitive with 4B-class models overall, while showing its advantages over models of comparable size in coding, mathematics, long-context understanding, tool use, and agentic tasks. font-family: -apple-system, BlinkMacSystemFont, "Segoe UI", Roboto…
English | MiniCPM5-2B-Math is a math and proof model built on MiniCPM5: a dense 2.5B model based on MiniCPM5-2B and post-trained for long-form mathematical reasoning and proof writing. It runs on a single consumer GPU or a laptop and keeps the native 128K context. - Strong for its size. As a plain single call with up to 128K tokens of thinking, 92.3 on AIME 2025, 89.8 on AIME 2026 and 81.8 on HMMT February 2026 (avg@16), and 100 / 93.3 / 90.9 with a 16-sample vote. The official general-purpose MiniCPM5-2B scores 87.1 / 89.8 / 68.4 under the same protocol. - Writes checkable proofs. On IMO-ProofBench Basic, an average single call with up to 128K tokens writes a complete proof for 21 of the…
A 2B-parameter Qwen3.5 fine-tune, part of the Opus-Distil line in the reaperdoesntknow open-weight portfolio. Text-generation / reasoning model, trained with Unsloth + Hugging Face TRL.