Qwen3.5 4B — Hardware Requirements & GPU Compatibility
VisionFunctions
Qwen3.5-4B is Alibaba's dense 4-billion-parameter vision-language model from the Qwen3.5 generation, sharing the same hybrid Gated DeltaNet and gated-attention architecture as its larger siblings. This is the post-trained, instruction-tuned release, able to process images together with text and handle general chat, coding assistance, and agentic tool use. It has a 262,144-token native context window, extensible up to roughly 1,010,000 tokens, and is released under the Apache 2.0 license. At around 4.7 billion parameters, 4-bit quantization needs under 3GB of memory, so it runs easily on almost any modern GPU or laptop.
Runs great on a 24 GB GPU
Runs from 8 GB (Q4_K_M · 128K)Best on 24 GB (Q4_K_M · ~201 tok/s · 256K)Macs from 8 GB (Q4_K_M · ~22 tok/s)
Score 98.5#64 of 106 open#12 of 36 that fit 24 GBApache 2.0
Reference GPU · RTX 4090 24 GB
Estimates, ±20% · how we estimate
llmrun Composite Score →
98.5
- 90% range 93.8–103.1
- #64 of 106 open models
- #12 of 36 that fit 24 GB
Run Qwen3.5 4B
Best fit for your RTX 5060. 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 for your RTX 5060
ollama run hf.co/unsloth/Qwen3.5-4B-GGUF:Q4_K_MQ4_K_M fits fully · 4.6 GB free · 16K context
Download: Qwen3.5-4B-Q4_K_M.gguf2.74 GBunsloth
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 3060 12GB (8 GB headroom) · Q4_K_M
- Generation speed
- ~72 tok/s
- generation speed
- Cost per 1M output tokens
- $0.23
- per 1M output tokens
How Much VRAM Does Qwen3.5 4B Need?
Select a quantization level; the listed hardware adheres to it.
| Quantization | VRAM | +32K | On your device | |
|---|---|---|---|---|
| 2.5 GB | 3 GB | 2-bit · K-quant — 2-bit quantization with K-quant improvements | Fits: 256K context · ~119 tok/s | |
| 2.5 GB | 3 GB | 3-bit small — 3-bit small quantization | Fits: 256K context · ~116 tok/s | |
| 2.7 GB | 3.2 GB | 3-bit medium — 3-bit medium quantization | Fits: 256K context · ~106 tok/s | |
| 2.8 GB | 3.3 GB | 4-bit · legacy — 4-bit legacy quantization | Fits: 256K context · ~104 tok/s | |
| 3.3 GB | 3.8 GB | 4-bit medium — 4-bit medium quantization — most popular sweet spot | Fits: 128K context · ~89 tok/s | |
| 3.8 GB | 4.3 GB | 5-bit medium — 5-bit medium quantization — good quality/size tradeoff | Fits: 128K context · ~77 tok/s | |
| 4.3 GB | 4.8 GB | 6-bit, very good — 6-bit quantization, very good quality | Fits: 128K context · ~68 tok/s | |
| 5.1 GB | 5.6 GB | 8-bit, near-lossless — 8-bit quantization, near-lossless | Fits: 128K context · ~57 tok/s |
Selected Q4_K_M
4-bit medium quantization3.4GBat 16K context
- Weights
- 2.8 GB
- KV cache
- 0.3 GB
- Overhead
- 0.3 GB
Fits RTX 5060 · 4.6 GB free · ~89 tok/s
≥50 tok/s needs ≥251 GB/s · RTX 5090 ~357, RTX 3090 Ti ~201
Which Hardware Can Run Qwen3.5 4B?
Q4_K_M · 3.3 GBQwen3.5 4B (Q4_K_M) requires 3.3 GB of VRAM to load the model weights. For comfortable inference with headroom for KV cache and system overhead, 5+ GB is recommended. Using the full 262K context window can add up to 5.2 GB, bringing total usage to 8.5 GB. 52 GPUs can run it, including NVIDIA GeForce RTX 5090, NVIDIA GeForce RTX 3090 Ti. 163 devices with unified memory can run Qwen3.5 4B, including Mac Studio (M3 Ultra, 256GB), Mac Studio (M3 Ultra, 512GB), MacBook Air 13" M3 (8 GB).
12 GB
12 GB tier →fits · 8.6 GB free · up to 256K ctx
10 GPUs · 1 device family
RTX 3080 Ti~182Simulate output on RTX 3080 TiRTX 5070~134Simulate output on RTX 5070RTX 4070~101Simulate output on RTX 4070iPhone~17Simulate output on Apple iPhone 17 Pro
By VRAM tier · typical GPUs and Macs at Q4_K_M
How good is Qwen3.5 4B?
The llmrun Score merges 3 public benchmark boards into a single figure, where 100 represents the average for open models.
Score vs VRAM in the 24 GB tier
- Qwen3.5 4B
- Other models
- Best Score for its VRAM
#12 of 36 that fit 24 GB8 of 36 fit your RTX 5060 (8 GB)
Each dot is a model at the best quant that fits a 24 GB GPU. The line steps up to the best Score available at each amount of VRAM; a dot below it is beaten by a model that needs no more memory.
Full 24 GB ranking →Benchmarks
Here are all 3 boards that have results for this model, sorted by their weight in the llmrun Score; badges show who ran the test.
- AIME 2024/2025
Epoch AI · Feb 2026Independent
Independent55.8%
#43/88 open
- MMLU-Pro
TIGER-LabIndependent
Independent79.1%
#26/146 open
- CritPt
Epoch AI · Feb 2026Lab-reported
Lab-reported0.0%
#81/84 open
In the 24 GB class
36 models fit 24 GBSame Score, only models that fit 24 GB at their best quant: the ranking a 24 GB owner actually chooses from.
- #1
Qwen3.8 27BAlibaba
122.0
~38 tok/s · 17.9 GB
- #2
Qwen3.6 27BAlibaba
111.2
~38 tok/s · 17.9 GB
- #3
Qwen3.6 35B A3BAlibaba
106.8
~151 tok/s · 22.0 GB
- #4
Gemma 4 31B ITGoogle
106.7
~32 tok/s · 21.7 GB
- #5
Gemma 4 26B A4B ITGoogle
105.8
~171 tok/s · 16.5 GB
- #6
Qwen3.5 35B A3BAlibaba
105.8
~151 tok/s · 22.0 GB
Alternatives to Qwen3.5 4B
Qwen 3.5 family · 34 models →Official models with an llmrun Score, each answering one question; numbers are at the best quant that fits a 24 GB GPU.
Other vendor
Gemma 3 4B IT
Google
- 34.5 lower Score
- 1.6 GB more VRAM
- 30 tok/s faster
Score 64.0 · 5 GB · ~231 tok/s
Similar models
8| Model | Score | VRAM | tok/s | Why it is similar |
|---|---|---|---|---|
| Qwen3.5 9BAlibaba | 99.4 | 6.6 GB | ~103 | similar score |
| WeVisOne VL 4BTencent | — | 4.7 GB | ~193 | same size |
| FrogNano 4B 2609Microsoft | — | 3.4 GB | ~201 | same size |
| Molmo2 4BAllen AI | — | 4.7 GB | ~193 | same size |
| Qwen3.6 27B MTPLX Optimized SpeedYoussofal | — | 4 GB | ~185 | same size |
| AFM 4.5BArcee AI | — | 4.3 GB | ~204 | same size |
| Xflux Text EncodersXLabs-AI | — | 12.7 GB | ~63 | same size |
| Qwen2.5 7B InstructAlibaba | 65.1 | 5.8 GB | ~131 | fits 24 GB |
Fine-tunes & variants(24)
- AREX Turbo
- Apodex 1.0 4B SFT
- CogEvol 4B
- Intern Decision 4B
- JOSIE 2 4B OSS
- JevK5
- Moe Expert Coder 4B
- NeoHorse 1 4B
- Prism 1 Pro
- Qwen3.5 4B Abliterated
- Qwen3.5 4B Claude 4.6 Opus Reasoning Distilled
- Qwen3.5 4B DFlash
- Qwen3.5 4B GLB 3D Coder
- Qwen3.5 4B PTBR
- Qwen3.5 4B Qwenputer
- Qwen3.5 4B Safety Thinking
- Qwen3.5 4B Super Coder
- Qwen3.8 4B
- Qwen3.8 4B Distill
- Qwen35 4B Soyuz Merged
- Qwopus3.5 4B Coder
- UserRepGen 4B
- Wald 4B
- Zira Researcher
Specifications
- Publisher
- Alibaba
- Family
- Qwen 3.5
- Parameters
- 4.7B
- Architecture
- Qwen3_5ForConditionalGeneration
- Context Length
- 262,144 tokens
- Vocabulary Size
- 248,320
- Release Date
- 2026-02-27
- License
- Apache 2.0
Frequently Asked Questions
How much VRAM does Qwen3.5 4B need?
Qwen3.5 4B requires 3.3 GB of VRAM at Q4_K_M, or 9.8 GB at BF16. Full 262K context adds up to 5.2 GB (8.5 GB total).
VRAM = Weights + KV Cache + Overhead
Weights = 4.7B × 4.8 bits ÷ 8 = 2.8 GB
KV Cache + Overhead ≈ 0.5 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.
KV Cache + Overhead ≈ 5.7 GB (at full 262K context)
VRAM usage by quantization
What's the best quantization for Qwen3.5 4B?
For Qwen3.5 4B, Q4_K_M (3.3 GB) offers the best balance of quality and VRAM usage. Q4_K_L (3.3 GB) provides better quality if you have the VRAM. The smallest option is IQ2_XXS at 1.8 GB.
VRAM requirement by quantization
★ Recommended — best balance of quality and VRAM usage.
Can I run Qwen3.5 4B on a Mac?
Yes — Mac Mini M2 (8 GB) and 145 other Macs can run Qwen3.5 4B. Apple Silicon uses unified memory, so the model shares RAM with the system. At Q4_K_M you need at least 3.3 GB of usable unified memory (RAM minus macOS overhead).
Can I run Qwen3.5 4B locally?
Yes — Qwen3.5 4B can run locally on consumer hardware. Popular tools include Ollama, LM Studio, and llama.cpp.
How fast is Qwen3.5 4B?
At Q4_K_M, Qwen3.5 4B can reach ~1472 tok/s on AMD Instinct MI350X. On NVIDIA GeForce RTX 4090: ~201 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 ÷ 3.26 × 0.65 = ~1595 tok/s
Estimated speed at Q4_K_M (3.3 GB)
Real-world results typically within ±20%. Speed depends on batch size, quantization kernel, and software stack.
What's the download size of Qwen3.5 4B?
At Q4_K_M, the download is about 2.80 GB. The full-precision BF16 version is 9.32 GB. The smallest option (IQ2_XXS) is 1.28 GB.
Which GPUs can run Qwen3.5 4B?
52 consumer GPUs can run Qwen3.5 4B at Q4_K_M. Top options include AMD Radeon RX 6700 XT, AMD Radeon RX 6800, AMD Radeon RX 6800 XT. 52 GPUs have plenty of headroom for comfortable inference.
Which devices can run Qwen3.5 4B?
163 devices with unified memory can run Qwen3.5 4B at Q4_K_M (3.3 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.