Vicuna 33B V1.3 — Hardware Requirements & GPU Compatibility
Chat
Vicuna-33B-v1.3 is LMSYS's chat assistant, fine-tuned from Meta's original LLaMA (33B) with supervised instruction tuning on around 125,000 user-shared conversations collected from ShareGPT.com. It was one of the models that popularized using GPT-4 and human preference judging as an evaluation method for open chat models, and is intended primarily for research on large language models and chatbots rather than production deployment. At 33 billion dense parameters it needs a high-end consumer GPU or multi-GPU setup once quantized. Context length is 2,048 tokens, inherited from the original LLaMA base. It is released under a non-commercial license, reflecting both LLaMA's original research-only terms and ShareGPT's usage terms, so it cannot be used commercially. It was published in June 2023; it is an early, now historical Vicuna release, later superseded by Vicuna v1.5 built on Llama 2.
Decent on a 24 GB GPU
Runs from 16 GB (Q3_K_S · 2K)Best on 24 GB (Q4_K_M · ~30 tok/s · 2K)Macs from 24 GB (Q4_K_S · ~11 tok/s)
Reference GPU · RTX 4090 24 GB
Estimates, ±20% · how we estimate
Run Vicuna 33B V1.3
One recommended way. Every other method, OS and quant is in the step-by-step guide.Each way opens a step-by-step guide for your OS and quant.
Recommended
ollama run hf.co/TheBloke/vicuna-33B-GGUF:Q4_K_MQ4_K_M fits fully on the RTX 4090 · 2.2 GB free · 2K context
Download: vicuna-33b.Q4_K_M.gguf19.62 GBTheBloke
Ways to runOther ways
each opens step-by-stepRun in cloud
Your own private copy · Q4_K_M · 24 GB GPU
- The easiest way to try this model
- Running in a few minutes, easy setup
- Chat right here or connect any app
$5+ bonus for new users
Runs on RTX 5090 (32 GB) (10 GB headroom) · Q4_K_M
- Generation speed
- ~54 tok/s
- generation speed
- Cost per 1M output tokens
- $2.58
- per 1M output tokens
How Much VRAM Does Vicuna 33B V1.3 Need?
Select a quantization level; the listed hardware adheres to it.
| Quantization | VRAM | +32K | On RTX 4090 (reference) | |
|---|---|---|---|---|
| 15.4 GB | — | 2-bit · K-quant — 2-bit quantization with K-quant improvements | Fits: 2K context · ~43 tok/s | |
| 15.9 GB | — | 3-bit small — 3-bit small quantization | Fits: 2K context · ~41 tok/s | |
| 17.7 GB | — | 3-bit medium — 3-bit medium quantization | Fits: 2K context · ~37 tok/s | |
| 18.1 GB | — | 4-bit · legacy — 4-bit legacy quantization | Fits: 2K context · ~36 tok/s | |
| 21.8 GB | — | 4-bit medium — 4-bit medium quantization — most popular sweet spot | Fits: 2K context · ~30 tok/s | |
| 25.9 GB | — | 5-bit medium — 5-bit medium quantization — good quality/size tradeoff | Does not fit: needs 25.9 GB | |
| 29.9 GB | — | 6-bit, very good — 6-bit quantization, very good quality | Does not fit: needs 30 GB | |
| 36.3 GB | — | 8-bit, near-lossless — 8-bit quantization, near-lossless | Does not fit: needs 36.3 GB |
Selected Q4_K_M
4-bit medium quantization21.8GBat 2K context
- Weights
- 20.9 GB
- KV cache
- 0.6 GB
- Overhead
- 0.3 GB
Fits RTX 4090 (reference) · 2.2 GB free · ~30 tok/s
≥50 tok/s needs ≥1.67 TB/s · RTX 5090 ~54
Or .
Which Hardware Can Run Vicuna 33B V1.3?
Q4_K_M · 21.8 GBVicuna 33B V1.3 (Q4_K_M) requires 21.8 GB of VRAM to load the model weights. For comfortable inference with headroom for KV cache and system overhead, 29+ GB is recommended. 7 GPUs can run it, including NVIDIA GeForce RTX 5090. 100 devices with unified memory can run Vicuna 33B V1.3, including Mac Studio (M3 Ultra, 256GB), Mac Studio (M3 Ultra, 512GB), Mac Studio M5 Max (36 GB).
20 GB
Q4_K_M · best fit Q3_K_L (18.6 GB)
1 GPU
RX 7900 XT~26Simulate output on RX 7900 XT32–40 GB
fits · 10.2 GB free · up to 2K ctx
1 GPU · 3 Mac classes · 4 device families
RTX 5090~54Simulate output on RTX 5090M5 Max36 GB~15Simulate output on Mac Studio M5 Max (36 GB)M532 GB~5Simulate output on MacBook Air 13" M5 (32 GB)Orin~6Simulate output on NVIDIA Jetson AGX Orin 32GBSnapdragon X Elite (X1E-80-100)~4Simulate output on Snapdragon X Elite Copilot+ PC
By VRAM tier · typical GPUs and Macs at Q4_K_M
How good is Vicuna 33B V1.3?
llmrun Score blends public benchmark boards into one number; 100 = the average open model.
Not enough data yet for an llmrun Score.
A Score needs results on enough independent boards. How the Composite Score is built
Benchmarks
This is the single board showing results for this model; badges show who ran the test.
- LMArena Text
LMArena · Oct 2026Arena
Arena1173.0
#159/188 open
Alternatives to Vicuna 33B V1.3
Vicuna family · 3 models →Similar models
6| Model | Score | VRAM | tok/s | Why it is similar |
|---|---|---|---|---|
| Nanbeige4.3 33BNanbeige | — | 23.7 GB | ~32 | same size |
| MiniCPM V 4.7 35B A3BOpenBMB | — | 22.1 GB | ~169 | same size |
| Llm Jp 4.1 32B A3b Thinkingllm-jp | — | 20.1 GB | ~188 | same size |
| Llm Jp 4.1 33B Thinkingllm-jp | — | 22.5 GB | ~35 | same size |
| Agnes 3.0 FlashAgnes-AI | — | 22.1 GB | ~35 | same size |
| NVIDIA Nemotron 3.5 Lightning 30B A3B BF16NVIDIA | — | 19.3 GB | ~197 | same size |
Specifications
- Publisher
- LMSYS
- Family
- Vicuna
- Parameters
- 33B
- Architecture
- LlamaForCausalLM
- Context Length
- 2,048 tokens
- Vocabulary Size
- 32,000
- Release Date
- 2023-06-21
Frequently Asked Questions
How much VRAM does Vicuna 33B V1.3 need?
Vicuna 33B V1.3 requires 21.8 GB of VRAM at Q4_K_M, or 72.6 GB at FP16.
VRAM = Weights + KV Cache + Overhead
Weights = 33B × 4.8 bits ÷ 8 = 19.8 GB
KV Cache + Overhead ≈ 2 GB (at 2K context + ~0.3 GB framework)
Fit ratings and hardware model lists check this model with room for a 16K-token context, which needs a little more memory.
VRAM usage by quantization
Can NVIDIA GeForce RTX 4090 run Vicuna 33B V1.3?
Yes, at Q5_0 (22.7 GB) or lower. Higher quantizations like Q5_K_S (25.0 GB) exceed the NVIDIA GeForce RTX 4090's 24 GB.
What's the best quantization for Vicuna 33B V1.3?
For Vicuna 33B V1.3, Q4_K_M (21.8 GB) offers the best balance of quality and VRAM usage. Q5_0 (22.7 GB) provides better quality if you have the VRAM. The smallest option is Q2_K at 15.4 GB.
VRAM requirement by quantization
★ Recommended — best balance of quality and VRAM usage.
Can I run Vicuna 33B V1.3 on a Mac?
Yes — Mac Mini M5 Pro (24 GB) and 106 other Macs can run Vicuna 33B V1.3, the smallest at Q4_K_S. Apple Silicon uses unified memory, so the model shares RAM with the system. At Q4_K_M you need at least 21.8 GB of usable unified memory (RAM minus macOS overhead), from Mac Studio M1 Max (32 GB).
Can I run Vicuna 33B V1.3 locally?
Yes — Vicuna 33B V1.3 can run locally on consumer hardware. Popular tools include Ollama, LM Studio, and llama.cpp.
How fast is Vicuna 33B V1.3?
At Q4_K_M, Vicuna 33B V1.3 can reach ~220 tok/s on AMD Instinct MI350X. On NVIDIA GeForce RTX 4090: ~30 tok/s. Speed depends mainly on GPU memory bandwidth. Real-world results typically within ±20%.
tok/s = (bandwidth GB/s ÷ model GB) × efficiency
Example: NVIDIA B200 → 8000 ÷ 21.8 × 0.65 = ~239 tok/s
Estimated speed at Q4_K_M (21.8 GB)
Real-world results typically within ±20%. Speed depends on batch size, quantization kernel, and software stack.
What's the download size of Vicuna 33B V1.3?
At Q4_K_M, the download is about 19.80 GB. The full-precision FP16 version is 66.00 GB. The smallest option (Q2_K) is 14.03 GB.
Which GPUs can run Vicuna 33B V1.3?
7 consumer GPUs can run Vicuna 33B V1.3 at Q4_K_M. Top options include AMD Radeon RX 7900 XTX, NVIDIA GeForce RTX 3090.
Which devices can run Vicuna 33B V1.3?
100 devices with unified memory can run Vicuna 33B V1.3 at Q4_K_M (21.8 GB), including AMD Ryzen AI 9 HX 370 (Strix Point) Laptop, ASUS Ascent GX10, Beelink GTR9 Pro (Ryzen AI Max+ 395, 128 GB), Framework Desktop (Ryzen AI Max+ 395, 128 GB). Apple Silicon Macs use unified memory shared between CPU and GPU, making them well-suited for local LLM inference.