★ Reading this for free? Get 25 structured AI courses + per-chapter AI tutor — the first chapter of every course free, no card.Start free in 30 secondsOr own every course: $149 once

WizardLM 13B: Evol-Instruct Model

Technical review of WizardLM 13B -- the model that pioneered Evol-Instruct training. Real benchmarks from the Open LLM Leaderboard, VRAM requirements, and honest 2026 assessment.

Released July 2023|Last updated March 13, 2026|By Local AI Master

Technical Specifications Overview

*Parameters: 13 billion
*Base Model: LLaMA 2 13B (v1.2) / LLaMA 1 13B (v1.0-v1.1)
*Context Window: 4,096 tokens
*Training: Evol-Instruct (evolved instruction tuning)
*Release: July 2023 (WizardLM team / Microsoft Research)
*License: Llama 2 Community License (v1.2)
*VRAM (Q4_K_M): ~8 GB
*Ollama: ollama run wizardlm:13b-llama2-q4_0

WizardLM 13B Evol-Instruct Architecture

WizardLM 13B architecture showing Evol-Instruct training pipeline: base LLaMA model fine-tuned with evolved instructions for improved instruction-following

👤
You
💻
Your ComputerAI Processing
👤
🌐
🏢
Cloud AI: You → Internet → Company Servers

Evol-Instruct: The Training Innovation Behind WizardLM

WizardLM's key contribution to the AI field is Evol-Instruct, a novel method for automatically generating high-complexity training data. Instead of relying on expensive human annotation (like InstructGPT) or distillation from proprietary models (like Alpaca), Evol-Instruct uses an LLM to evolve simple instructions into progressively more complex ones.

How Evol-Instruct Works

The method uses two evolution strategies described in arXiv:2304.12244:

  • In-Depth Evolution: Makes instructions harder by adding constraints, increasing reasoning steps, concretizing abstract concepts, or requiring multi-step solutions
  • In-Breadth Evolution: Generates entirely new instructions inspired by existing ones, expanding topic coverage and diversity

In-Depth Evolution Example

Original instruction:
"Write a function to sort a list"
↓ Evolved ↓
Evolved instruction:
"Write a function to sort a list of dictionaries by multiple keys, handling missing keys gracefully, with O(n log n) time complexity and stable ordering"

In-Breadth Evolution Example

Original instruction:
"Explain photosynthesis"
↓ Evolved ↓
New related instruction:
"Compare aerobic and anaerobic cellular respiration, including the role of the electron transport chain"

The WizardLM team evolved 52K Alpaca instructions through multiple rounds of Evol-Instruct, producing the training dataset used to fine-tune the LLaMA base model. This approach proved effective: on the Evol-Instruct testset, WizardLM 7B showed competitive performance with ChatGPT on complex instructions while using a fraction of the parameters. The same recipe was scaled up to WizardLM 30B, which trades roughly 2x the VRAM for a modest quality gain over this 13B release.

Historical note: The Evol-Instruct methodology influenced many subsequent models and training approaches. The WizardLM team later applied similar techniques to create WizardCoder and WizardMath, demonstrating the generalizability of evolved instruction tuning. The community also blended the method with Vicuna's conversational data in WizardVicuna, best known today through its uncensored builds (see our guide to uncensored local LLMs).

Real Benchmark Performance

All benchmark scores below are from the HuggingFace Open LLM Leaderboard -- not marketing claims.

MMLU: 13B Model Comparison

MMLU Score (%) - Source: Open LLM Leaderboard

Vicuna 13B v1.556.7 Score (%)
56.7
Llama 2 13B Chat54.6 Score (%)
54.6
WizardLM 13B v1.254.6 Score (%)
54.6
Nous Hermes 13B50.4 Score (%)
50.4
CodeLlama 13B32.8 Score (%)
32.8

HellaSwag & ARC Comparison

HellaSwag & ARC-Challenge (%) - Source: Open LLM Leaderboard

WizardLM 13B HellaSwag82.2 Score (%)
82.2
Llama 2 13B HellaSwag81 Score (%)
81
WizardLM 13B ARC59 Score (%)
59
Llama 2 13B ARC58.1 Score (%)
58.1

Multi-dimensional Benchmark Analysis

Performance Metrics

MMLU
54.6
HellaSwag
82.2
ARC-Challenge
59
TruthfulQA
47.3
Winogrande
71.9

Benchmark Context

WizardLM 13B v1.2 performs comparably to other 13B-class models from the same era (mid-2023). Its MMLU of 54.6% places it in the middle of the 13B pack. The model's real strength is instruction-following quality -- the Evol-Instruct training makes it handle complex multi-step instructions better than raw benchmark scores suggest. However, standard benchmarks show it barely moves its base model (Llama 2 13B, 54.3% MMLU) on knowledge tasks.

Note: all scores are from the original Open LLM Leaderboard (v1) result files on Hugging Face, which list the Llama 2 13B base as the HellaSwag and ARC comparison. This site has not run these benchmarks.

VRAM Requirements by Quantization

VRAM Usage by Quantization Level

Download sizes for the Llama 2-based WizardLM 13B on Ollama's wizardlm tag list (checked November 2, 2026): wizardlm:13b-llama2-q4_0 7.4 GB, -q4_K_M 7.9 GB, -q5_K_M 9.2 GB, -q8_0 13.8 GB, -fp16 26 GB. Add about 1 GB for the KV cache and runtime at the 4K context.

Q4_K_M

~8 GB

4-bit quantization, minimal quality loss

Best for: RTX 3060 12GB, M1/M2 16GB

Q5_K_M

~9.2 GB

5-bit quantization, good quality balance

Best for: RTX 4060 Ti 16GB, M2 Pro 16GB

Q8_0

~14 GB

8-bit quantization, near-lossless

Best for: RTX 4070 Ti, M2 Pro 32GB

FP16

~26 GB

Full precision, no quality loss

Best for: RTX 3090/4090, A5000

Installation & Setup Guide

System Requirements

System Requirements

▸
Operating System
Windows 10/11, macOS 12+, Ubuntu 20.04+, Linux
▸
RAM
16GB minimum (8GB for Q4 quantization with Ollama)
▸
Storage
8GB for Q4_K_M model, 26GB for FP16
▸
GPU
Optional: RTX 3060+ (12GB VRAM) or Apple M1/M2
▸
CPU
8+ cores recommended (Intel i7/AMD Ryzen 7+)
1

Install Ollama

Install the Ollama runtime for local model deployment

$ curl -fsSL https://ollama.com/install.sh | sh
2

Download WizardLM 13B

Download the Llama 2-based WizardLM 13B at Q4_0 (7.4GB download, needs ~8GB VRAM). Ollama has no plain wizardlm:13b tag

$ ollama pull wizardlm:13b-llama2-q4_0
3

Run WizardLM 13B

Start an interactive chat session with WizardLM 13B

$ ollama run wizardlm:13b-llama2-q4_0
4

Test instruction following

Test the model with a structured instruction-following task

$ ollama run wizardlm:13b-llama2-q4_0 "Explain the difference between depth-first and breadth-first search with pseudocode"

Terminal Demo

Terminal
$ollama run wizardlm:13b-llama2-q4_0 "Explain how a hash table works, including collision resolution strategies"
A hash table is a data structure that maps keys to values using a hash function. **How it works:** 1. A hash function converts a key into an array index 2. The value is stored at that computed index 3. Lookup is O(1) average case **Collision Resolution Strategies:** **1. Chaining (Separate Chaining)** - Each array slot holds a linked list - Colliding elements are appended to the list - Simple to implement, degrades gracefully **2. Open Addressing** - All elements stored in the array itself - On collision, probe for next empty slot - Variants: Linear probing, Quadratic probing, Double hashing **3. Robin Hood Hashing** - Variant of open addressing - Elements that are "richer" (closer to their ideal position) give up their slot to "poorer" elements - Reduces variance in probe lengths **Trade-offs:** - Chaining: Better with high load factors, uses more memory - Open addressing: Better cache performance, sensitive to load factor - Load factor should stay below 0.75 for good performance
$_

Comparison with Other 13B Models

Only comparing local 13B-class models -- cloud models (ChatGPT, Claude) operate at completely different scales and are not meaningful comparisons.

ModelSizeRAM RequiredSpeed ceiling (RTX 3060 12GB)MMLU (Open LLM Leaderboard v1)Licence
WizardLM 13B v1.27.4GB (Q4)~8GB VRAM~49 tok/s ceiling54.6%Free (Llama 2 License)
Llama 2 13B Chat7.4GB (Q4)~8GB VRAM~49 tok/s ceiling54.6%Free (Llama 2 License)
Vicuna 13B v1.57.4GB (Q4)~8GB VRAM~49 tok/s ceiling56.7%Free (Llama 2 License)
Nous Hermes 13B7.4GB (Q4)~8GB VRAM~49 tok/s ceiling50.4%Free (LLaMA 1 based)
CodeLlama 13B7.4GB (Q4)~8GB VRAM~49 tok/s ceiling32.8%Free (Llama 2 License)

Speed is an upper bound, not a measurement: an RTX 3060 12GB's 360 GB/s of memory bandwidth divided by the 7.4 GB Q4_0 file. All five share the Llama 13B layout, so the ceiling is the same; real speed lands below it.

Key Takeaways

  • *All 13B models share similar VRAM footprints (~8GB at Q4_K_M) since they share the LLaMA 2 architecture
  • *WizardLM 13B (54.6% MMLU) sits just behind Vicuna 13B v1.5 (56.7%); its case rests on instruction-following from Evol-Instruct, not on knowledge scores
  • *Llama 2 13B Chat lands on the same MMLU (54.6%): both are fine-tunes of the same base, and instruction tuning adds little knowledge
  • *CodeLlama 13B scores far lower on MMLU (32.8%) because it was specialized for code, not general knowledge

Practical Use Cases

Instruction Following

WizardLM 13B excels at structured multi-step instructions thanks to Evol-Instruct training. Good for drafting documents with specific formatting, step-by-step explanations, and constrained writing tasks.

"Write a 3-paragraph summary of X with a formal tone, including statistics..."

Local Privacy

All inference runs on your own hardware. No data leaves your machine. Useful for processing sensitive documents, internal communications, or proprietary information without cloud API exposure.

100% offline operation, no API calls

Learning & Experimentation

At 13B parameters, the model is small enough for experimentation. Good for learning about LLM behavior, prompt engineering practice, and understanding Evol-Instruct training effects.

Runs on consumer hardware (16GB RAM)

What WizardLM 13B Is NOT Good For

  • *Factual accuracy: With 54.6% MMLU, the model hallucinates frequently on knowledge-intensive tasks. Do not rely on it for factual claims without verification.
  • *Production deployment: Newer models like Llama 3 8B (~66% MMLU) and Mistral 7B v0.3 (~63% MMLU) outperform WizardLM 13B while using less VRAM.
  • *Code generation: Dedicated coding models like CodeLlama 13B or DeepSeek Coder 6.7B are better for programming tasks.
  • *Long context: The 4,096 token context window is limiting by 2026 standards. Newer models offer 8K-128K context.

Honest 2026 Assessment

WizardLM 13B was released in July 2023 and is nearly three years old. The local AI landscape has changed dramatically since then. Here is an honest evaluation of where it stands in 2026.

Still Relevant For

  • * Understanding Evol-Instruct training methodology
  • * Running on older/limited hardware (8GB VRAM is accessible)
  • * Historical comparison and benchmarking reference
  • * Learning about instruction-tuned models
  • * Offline privacy-focused use where quality is not critical

Better Alternatives in 2026

  • * Llama 3 8B -- ~66% MMLU, 8K context, ~6GB VRAM
  • * Mistral 7B v0.3 -- ~63% MMLU, 32K context, ~6GB VRAM
  • * Phi-3 Mini 3.8B -- ~71% MMLU, 128K context, ~3GB VRAM
  • * Gemma 2 9B -- ~71% MMLU, 8K context, ~7GB VRAM
  • * Qwen 2.5 7B -- ~74% MMLU, 128K context, ~6GB VRAM

Bottom Line

WizardLM 13B is historically significant as the model that popularized Evol-Instruct training. For new deployments in 2026, smaller and newer models deliver substantially better performance per VRAM. If you are starting fresh, consider Llama 3 8B or Mistral 7B instead. If you are already using WizardLM 13B and it meets your needs, there is no urgent reason to switch, but upgrading will yield meaningfully better results.

Reading now
Join the discussion

Ready to Go Beyond Tutorials?

25 structured courses with hands-on chapters - build RAG chatbots, AI agents, and ML pipelines on your own hardware.

Frequently Asked Questions

What are WizardLM 13B's actual benchmark scores?

On the original HuggingFace Open LLM Leaderboard, WizardLM 13B v1.2 scored 54.6% on MMLU, 82.2% on HellaSwag and 59.0% on ARC-Challenge. That puts it level with Llama 2 13B Chat (54.6% MMLU) and just behind Vicuna 13B v1.5 (56.7% MMLU) in the 13B parameter class.

How much VRAM does WizardLM 13B need?

WizardLM 13B needs approximately 8 GB VRAM at Q4_K_M quantization, 9-10 GB at Q5_K_M, 14 GB at Q8_0, and 26 GB at full FP16 precision. A 16GB GPU (RTX 4060 Ti 16GB) or Apple M1/M2 with 16GB unified memory can run Q4_K_M or Q5_K_M comfortably. Install via Ollama: ollama run wizardlm:13b-llama2-q4_0 (there is no plain wizardlm:13b tag).

What is Evol-Instruct and how does it train WizardLM?

Evol-Instruct is a novel training methodology introduced in the WizardLM paper (arXiv:2304.12244). It evolves simple instructions into increasingly complex ones through two strategies: in-depth evolution (adding constraints, deepening reasoning, concretizing) and in-breadth evolution (generating new related topics). This creates diverse, high-complexity training data without expensive human annotation.

Is WizardLM 13B still worth using in 2026?

WizardLM 13B is a 2023 model that has been surpassed by newer alternatives. For instruction-following, consider Llama 3 8B (~66% MMLU, 8GB VRAM) or Mistral 7B v0.3 (~63% MMLU, 6GB VRAM), both of which outperform WizardLM 13B while using less VRAM. WizardLM 13B remains historically significant for pioneering the Evol-Instruct methodology.

What license does WizardLM 13B use?

WizardLM 13B v1.2 uses the Llama 2 Community License from Meta, which permits commercial use provided you comply with Meta's acceptable use policy and have fewer than 700 million monthly active users. Earlier versions (v1.0, v1.1) used the original LLaMA license which was more restrictive.

🎯
AI Learning Path

Go from reading about AI to building with AI

25 structured courses. Hands-on projects. Runs on your machine. Start free.

Or own it for life — Lifetime $149 $599, pay once
Barsha Mohapatra

Written by

Barsha Mohapatra

Founder and Editor, Local AI Master · Founder, Little AI Master

Founder of Little AI Master and the editor behind every course and guide on Local AI Master. Years of tutoring students taught her where AI education actually fails.

📅 Published: September 28, 2025🔄 Last Updated: November 2, 2026✓ Manually Reviewed

Related Guides

Continue your local AI journey with these comprehensive guides

More on Local AI Hardware
See the full AI Hardware Guide 2026 guide.
📚
Free · no account required

Grab the AI Starter Kit — career roadmap, cheat sheet, setup guide

No spam. Unsubscribe with one click.

🎯
AI Learning Path

Found your model? Now build something with it.

25 hands-on courses — RAG, agents, fine-tuning — all running locally. First chapter free, no card.

Or own it for life — Lifetime $149 $599, pay once
Free Tools & Calculators