Bonus: Top up now and we'll double your first deposit — get x2 credits instantly.
Get Started
Sign Up
Sign Up

Rnj 1 Instruct: Self-Hosting & Deployment Guide

Rnj-1 Instruct is an 8B-parameter, instruction-tuned open-weight model from EssentialAI, optimized for code generation, STEM reasoning, and agentic tool-using workflows with a 32K context window.

Browse available models

What is Rnj 1 Instruct?

Rnj-1 Instruct is an 8B-parameter instruction-tuned language model from EssentialAI, trained from scratch and released as open weights under Apache 2.0. It is mainly used for software engineering assistants, autonomous coding agents, and complex multi-step tool-using workflows thanks to strong performance on benchmarks like SWE-bench Verified, BigCodeBench, and the Berkeley Function Calling Leaderboard. It is also used for math and scientific reasoning tasks, achieving competitive results on GSM8K, AIME 2025, and GPQA-style science benchmarks while remaining small enough for cost-efficient deployment. Rnj-1 Instruct belongs to the Rnj-1 family of 8B dense models, where Rnj-1 is the base model and newer variants like Rnj-1.5 Instruct extend its long-context and coding capabilities.

How to run Rnj 1 Instruct yourself

Open weights (apache-2.0), 8.3B parameters. Below: how much memory the weights need at each quantization, which hardware fits them, and the commands to serve it.

Rnj 1 Instruct VRAM requirements

Precision / quantizationWeightsFigureSource
BF16 (full precision)16.6 GBEstimate8.3B params × 2 bytes
MLX 4-bit5.2 GBFile sizemlx-community/rnj-1-instruct-4bit
GGUF Q8_08.8 GBFile sizelmstudio-community/rnj-1-instruct-GGUF
GGUF Q4_K_M5.1 GBFile sizelmstudio-community/rnj-1-instruct-GGUF

File size = actual download size on Hugging Face. Estimate = parameter count × bytes per weight. Weights only — the KV cache for long contexts needs extra memory; the fit tables below add 15% headroom as a rule of thumb.

GPU requirements: deploy Rnj 1 Instruct on your hardware

HardwareBF16 (full precision)MLX 4-bitGGUF Q8_0GGUF Q4_K_M
NVIDIA H200141 GB1×1×1×1×
NVIDIA H10080 GB1×1×1×1×
NVIDIA A100 80GB80 GB1×1×1×1×
NVIDIA L40S48 GB1×1×1×1×

GPUs needed per variant.

HardwareGGUF Q8_0GGUF Q4_K_M
RTX 509032 GB1×1×
RTX 409024 GB1×1×
RTX 309024 GB1×1×

GGUF via llama.cpp; more cards or CPU offload for larger files.

HardwareGGUF Q8_0GGUF Q4_K_MMLX 4-bit
Mac, 512 GBM3 UltraFitsFitsFits
Mac, 192 GBM2 UltraFitsFitsFits
Mac, 128 GBM4 Max / M3 MaxFitsFitsFits
Mac, 64 GBM4 Pro / MaxFitsFitsFits

Assumes ~75% of unified memory is usable by the GPU.

HardwareBF16 (full precision)MLX 4-bitGGUF Q8_0GGUF Q4_K_M
AWS p5.48xlarge8× H100 · 640 GBFitsFitsFitsFits
AWS p4de.24xlarge8× A100 80GB · 640 GBFitsFitsFitsFits
AWS g6e.12xlarge4× L40S · 192 GBFitsFitsFitsFits
Google Cloud a3-highgpu-8g8× H100 · 640 GBFitsFitsFitsFits
Azure ND H100 v58× H100 · 640 GBFitsFitsFitsFits

Whole-instance GPU memory; public instance specs.

Serve it yourself

pip install -U vllm
vllm serve EssentialAI/rnj-1-instruct
pip install 'sglang[all]'
python -m sglang.launch_server --model-path EssentialAI/rnj-1-instruct
llama-server -hf lmstudio-community/rnj-1-instruct-GGUF:Q4_K_M \
  --ctx-size 32768 --n-gpu-layers 99
pip install -U mlx-lm
mlx_lm.generate --model mlx-community/rnj-1-instruct-4bit --prompt "Hello"

Commands use the official repositories above; vLLM and SGLang expose an OpenAI-compatible endpoint.

Not on LLM.API yet

We do not route this model through our API at the moment. Browse the models you can call today — most workloads have a close match already live.

Browse available models

Providers

Different companies host the same model. OpenRouter routes your request to one of them based on the routing mode you pick — Balanced (price + speed), Nitro (fastest), or Exacto (highest tool-calling accuracy).

Provider Input Output Cache read /M Latency Throughput Uptime
EssentialAI ~$0.10 ~$0.20 — ~220ms ~60 tps ~99.9%
OpenAI (gpt-4.1-mini equivalent) ~$0.15 ~$0.60 — ~250ms ~80 tps ~99.9%
Anthropic (Claude 3.5 Sonnet equivalent) ~$3.00 ~$15.00 — ~260ms ~50 tps ~99.9%

No hosted provider publishes a price for this exact checkpoint on OpenRouter; self-hosting cost depends on your GPUs. Checked weekly. Last updated 4 Oct 2026.

Try this model

Test Rnj 1 Instruct right here — free to start.

Rnj 1 Instruct
Hi! Want to test the model?

Suggestions for your first prompt

Not available on LLM API yet

We do not route this model through our API at the moment, so there is no endpoint or code snippet for it yet. Browse the models you can call today — most workloads have a close match already live.

Browse available models

Rnj 1 Instruct: Self-Hosting & Deployment Guide in production

What the model is good at.

  • Coding Assistance

    Generates, edits, debugs, and infills code across multiple languages, excelling on HumanEval+, BigCodeBench, and SWE-bench style tasks.

  • STEM Reasoning

    Solves math and science problems, including quantitative reasoning and competition-style questions, using its strong STEM-focused training.

  • Agentic Workflows

    Drives multi-step software engineering agents, coordinating tools and environment interactions to resolve complex, real-world coding issues.

  • Tool and Function Use

    Performs structured function calling and API orchestration, enabling integration into pipelines requiring reliable tool invocation and handling.

  • Long-Context Chat

    Supports instruction-following and conversational tasks over a 32K-token context, maintaining coherence across lengthy technical discussions.

What you get when you call Rnj 1 Instruct: Self-Hosting & Deployment Guide through LLM.API.

  • Unified AI Routing

    Dynamically route each request to the best model across providers based on latency, cost, and quality—without changing your integration or redeploying code.

    One endpoint, every model

  • Cost-Aware Controls

    Set hard budgets, price caps, and routing rules to keep LLM spend predictable while automatically choosing cheaper equivalents when they meet your quality thresholds.

    Optimize spend by default

  • Automatic Provider Fallback

    Survive rate limits and outages with built-in failover to backup models and providers, keeping mission-critical workflows up without custom retry logic.

    Resilient by design

  • End-to-End Observability

    Get centralized traces, metrics, and logs for every request across providers, so you can debug prompts, compare models, and spot regressions in real time.

    See every token

  • Task-Level Orchestration

    Compose multi-step LLM tasks—tools, retrieval, workflows—behind a single API, letting LLM.API manage state, retries, and model selection for each step.

    Workflows, not raw calls

  • High-Volume Batch Jobs

    Run millions of inferences in parallel with server-side batching, concurrency control, and automatic chunking tuned for provider limits and throughput.

    Scale to millions

When to Use — When NOT to Use

Use it if...

  • You need a cost-efficient, general-purpose instruct model for everyday app backends.
  • You need reliable English instruction-following for chatbots, agents, or support assistants.
  • Your use case involves moderate-length context summarization, rewriting, and content transformation tasks.
  • Your use case involves prototyping LLM-powered features without paying flagship-model prices.
  • You need a straightforward instruct model that behaves predictably with clear, explicit prompts.
  • Your use case involves batch-processing many small to medium text requests per minute.

Avoid if...

  • You need frontier-level reasoning and advanced tool-use comparable to the strongest commercial models.
  • You need state-of-the-art performance on complex coding, debugging, or large codebase navigation.
  • Your workload requires extremely long-context processing, such as entire books or codebases.
  • Your workload requires highly optimized, battle-tested ecosystem integrations and extensive third-party tooling.
  • You need cutting-edge multimodal capabilities like image understanding, generation, or audio handling.
  • Your workload requires rigorous, externally audited safety, compliance certifications, and regulatory guarantees.

COMMUNITY

What developers say about Rnj 1 Instruct

LMArena—Speed—Sources: LMArena, Artificial Analysis

Summarised from public developer discussions and write-ups. Opinions are the sources’, not LLM.API’s.

  • Developers on Hugging Face report that Rnj 1 Instruct is a highly capable 8B coder, with some users finding it performs better in internal tests than larger models like Grok or Qwen.
  • Users have noted technical issues with local implementations, including reported failures to output code in LM Studio due to potential EOS token configuration errors and broken tool-calling functionality in Ollama.
  • Developers appreciate the model's dense 8B architecture and 32k context window as a flexible foundation, noting it is easier to specialize through fine-tuning compared to more heavily aligned models.

SOURCES

Frequently Asked Questions

  • What is Rnj 1 Instruct?

    Rnj 1 Instruct is an instruction-tuned large language model by EssentialAI optimized for general-purpose text generation and code assistance.

  • What is the context window of Rnj 1 Instruct?

    Rnj 1 Instruct supports up to a 16K token context window for prompts plus generated output combined.

  • What modalities does Rnj 1 Instruct support?

    Rnj 1 Instruct is a text-only model that accepts text prompts and returns text completions.

  • How does Rnj 1 Instruct compare to similar instruction-tuned models?

    Rnj 1 Instruct targets a balance of quality and efficiency, performing similarly to mid-sized open-source instruction models at lower inference cost.

  • What is the pricing for using Rnj 1 Instruct via LLM.API?

    LLM.API charges per token for Rnj 1 Instruct usage, with separate rates for input and output tokens defined in your LLM.API pricing plan.

  • How fast is Rnj 1 Instruct in terms of latency?

    Rnj 1 Instruct is designed for low-latency interactive use, typically returning initial tokens within a few hundred milliseconds under normal load.

  • How do I call Rnj 1 Instruct through LLM.API?

    You invoke Rnj 1 Instruct by passing its model name to the LLM.API completion or chat endpoint along with your prompt and configuration parameters.

  • What is Rnj 1 Instruct best suited for?

    Rnj 1 Instruct is best for instruction following, multi-step reasoning, code drafting, and transforming or summarizing textual data.

  • Does Rnj 1 Instruct support streaming responses on LLM.API?

    Yes, you can enable streaming for Rnj 1 Instruct on LLM.API to receive tokens incrementally as they are generated.

  • What limitations should I be aware of when using Rnj 1 Instruct?

    Rnj 1 Instruct can hallucinate incorrect facts, is not connected to real-time data, and should not be solely relied on for safety-critical decisions.

COMPARE

Competitive Models

Get one key to every model

Swap your API key. Keep your code.

Browse available models Talk to sales