8.7B- Context
- 33Ktokens
- Released
- 11mo ago
- Size
- 8.3Bparameters
Rnj-1 Instruct is Essential AI's instruction-tuned version of its Rnj-1 base model, an open-weight dense model the company trained from scratch and announced as its first contribution to open-source AI. The maker's launch announcement claims it outperforms similarly sized open-weight models on agentic coding, its stated target ability. The model card says it is built for code, math, and science work, including tool calling and use within agentic frameworks such as mini-SWE-agent.
How it compares
Additional source evidence
- Source model
- essentialai/rnj-1-instruct
- Source family
- rnj
- Last updated
- 2025-12-04
- Model context
- 33K
- Model output limit
- 33K
- Open weights
- Yes
- File attachments
- No
- Reasoning
- No
- Tool calling
- Yes
- Temperature control
- Yes
Downloadable weights
Complete source record (JSON)
{
"id": "essentialai/rnj-1-instruct",
"name": "Rnj-1 Instruct",
"description": "Open 8B Essential AI model for code, STEM reasoning, and agentic tool use",
"family": "rnj",
"last_updated": "2025-12-04",
"attachment": false,
"reasoning": false,
"tool_call": true,
"temperature": true,
"modalities": {
"input": [
"text"
],
"output": [
"text"
]
},
"limit": {
"context": 32768,
"output": 32768
},
"release_date": "2025-12-04",
"open_weights": true,
"license": "Apache 2.0",
"weights": [
{
"label": "Hugging Face",
"url": "https://huggingface.co/EssentialAI/rnj-1-instruct"
}
]
}Frequently asked questions
Running locally
Rnj-1 Instruct needs an estimated 6.3 GB of VRAM at Q4, 10.7 GB at Q8 and 19.6 GB at FP16. All figures are estimates from the 8.3B parameter count, covering weights plus about 15% and 0.5 GB for a short context.
Yes. A 24 GB GPU fits Rnj-1 Instruct at FP16 (estimated 19.6 GB), 16 GB and 12 GB GPUs fit at Q8 (estimated 10.7 GB), and an 8 GB GPU fits at Q4 (estimated 6.3 GB).
Yes. At Q4 Rnj-1 Instruct needs an estimated 6.3 GB, and assuming macOS lets the GPU use about 75% of unified memory, the smallest common Mac memory size that fits is an estimated 16 GB.
Privacy and data
Yes. Rnj-1 Instruct has open weights, so you can self-host it for full control of your data.
Capabilities and specs
The context window is 33K tokens (32,768) and the maximum output is 33K tokens (32,768).
No. Rnj-1 Instruct only accepts text input, so it cannot read images, audio, video or PDFs.
Yes. Rnj-1 Instruct supports tool calling (function calling).
No. Rnj-1 Instruct does not have a reasoning or thinking mode.
Quality and benchmarks
The closest alternatives, matched on model kind and release date (no comparable price), are Molmo 2 8B by Ai2, Olmo 3.1 32B Instruct by Ai2 and ERNIE 4.5 VL 28B A3B Thinking by Baidu.
Molmo 2 8BRnj-1 Instruct vs Molmo 2 8BOlmo 3.1 32B InstructRnj-1 Instruct vs Olmo 3.1 32B InstructERNIE 4.5 VL 28B A3B ThinkingRnj-1 Instruct vs ERNIE 4.5 VL 28B A3B Thinking
Licensing and access
Rnj-1 Instruct has open weights, released under the Apache 2.0 license.
Yes. Rnj-1 Instruct is licensed under Apache 2.0, a permissive license that allows commercial use.
From the Hugging Face repository EssentialAI/rnj-1-instruct.
Basics
Rnj-1 Instruct was made by Essential AI and released on December 4, 2025.
Yes. Rnj-1 Instruct has open weights.
Generated from Vertical Tabs data, updated November 10, 2026.
Similar models
8.7B
32.2BOlmo 3.1 32B Instruct
29.7BERNIE 4.5 VL 28B A3B Thinking
AutoGLM Phone 9B MultilingualAutoGLM Phone 9B Multilingual
959MPaddleOCR-VL
GLM-4.6V-FlashGLM-4.6V-Flash
26.1BTrinity Mini
6.1BTrinity Nano Preview
24BLFM2 24B A2B
14BMinistral 14B
14BMinistral 3 14B
8B
Discussion
No comments yet
Be the first to share your thoughts.