Skip to content
Weights Not Walls
Menu

Qwen

Qwen3.5-4B

Reviewedaccepts images
Size
4.7B parameterscounted from the safetensors header on the Hub
Licence
custom, read it, Apache License 2.0Our reading, check the licence: A custom licence we have not read for this model: other makers use the same name for different texts. Read the model's own licence file before you use it.Licence text
On the Hub since
27 Feb 2026the date the repo was created
Context length
256k tokensfrom the model's config
Fine-tune
of Qwen/Qwen3.5-4B-BaseQwen3_5ForConditionalGeneration
On Hugging Face
Qwen/Qwen3.5-4B4.6M downloads in the last 30 days, 1.3k likes

Scores

2 independent results on 2 eval sets
OTIS Mock AIME 2024-2025Re-run by Epoch AI
  • effort none55.8%±6.228 Aug 2026, log

Epoch's name: qwen3.5-4b. Best closed model on this set: claude-fable-5_high, 100.0% (Anthropic). All 318 results

CritPtReported

Epoch's name: qwen3.5-4b. Best closed model on this set: gemini-4-carbon_high, 36.0% (Google DeepMind). All 271 results

Claimed by the maker Claimed

The model card has no claimed scores in a machine-readable form (the card's model-index), so none are shown. The maker may give scores in the card text: read the card.

Other views, which we link to and do not copy: look this model up on Artificial Analysis and the Arena leaderboard.

Independent scores from Epoch AI (CC-BY 4.0), fetched 19 Nov 2026: Epoch AI, ‘Capabilities & benchmarking’. Published online at epoch.ai. Retrieved from ‘https://epoch.ai/benchmarks’ [online resource]. Each row links its own source. What the badges mean.

Run it yourself

file sizes read from the Hub 19 Nov 2026

Memory needed = weights + context cache + 0.50 GB for the runtime. Context cache at 8k tokens: 0.25 GB, worked out from the model's config (Qwen/Qwen3.5-4B), at 16-bit precision. Choose a machine above to see a verdict for each format. Method

GGUF for llama.cpp, Ollama, LM Studio and others

unsloth/Qwen3.5-4B-GGUF (converted by unsloth), 1.7M downloads. Sizes are the real files on the Hub, all parts added up. For image input you also need the projector file mmproj-F16.gguf (0.63 GB), not counted below.

FormatSizeMemory neededGet it
UD-IQ3_XXS1.8 GBfile size2.6 GB1.8 + 0.25 + 0.50
Commands
llama.cpp
llama-server -hf unsloth/Qwen3.5-4B-GGUF:UD-IQ3_XXS
Ollama
ollama run hf.co/unsloth/Qwen3.5-4B-GGUF:UD-IQ3_XXS
Download
hf download unsloth/Qwen3.5-4B-GGUF --include "Qwen3.5-4B-UD-IQ3_XXS.gguf" --local-dir Qwen3.5-4B-GGUF
Q3_K_S2.0 GBfile size2.7 GB2.0 + 0.25 + 0.50
Commands
llama.cpp
llama-server -hf unsloth/Qwen3.5-4B-GGUF:Q3_K_S
Ollama
ollama run hf.co/unsloth/Qwen3.5-4B-GGUF:Q3_K_S
Download
hf download unsloth/Qwen3.5-4B-GGUF --include "Qwen3.5-4B-Q3_K_S.gguf" --local-dir Qwen3.5-4B-GGUF
Q3_K_M2.1 GBfile size2.9 GB2.1 + 0.25 + 0.50
Commands
llama.cpp
llama-server -hf unsloth/Qwen3.5-4B-GGUF:Q3_K_M
Ollama
ollama run hf.co/unsloth/Qwen3.5-4B-GGUF:Q3_K_M
Download
hf download unsloth/Qwen3.5-4B-GGUF --include "Qwen3.5-4B-Q3_K_M.gguf" --local-dir Qwen3.5-4B-GGUF
UD-Q3_K_XL2.3 GBfile size3.0 GB2.3 + 0.25 + 0.50
Commands
llama.cpp
llama-server -hf unsloth/Qwen3.5-4B-GGUF:UD-Q3_K_XL
Ollama
ollama run hf.co/unsloth/Qwen3.5-4B-GGUF:UD-Q3_K_XL
Download
hf download unsloth/Qwen3.5-4B-GGUF --include "Qwen3.5-4B-UD-Q3_K_XL.gguf" --local-dir Qwen3.5-4B-GGUF
IQ4_XS2.3 GBfile size3.1 GB2.3 + 0.25 + 0.50
Commands
llama.cpp
llama-server -hf unsloth/Qwen3.5-4B-GGUF:IQ4_XS
Ollama
ollama run hf.co/unsloth/Qwen3.5-4B-GGUF:IQ4_XS
Download
hf download unsloth/Qwen3.5-4B-GGUF --include "Qwen3.5-4B-IQ4_XS.gguf" --local-dir Qwen3.5-4B-GGUF
IQ4_NL2.4 GBfile size3.2 GB2.4 + 0.25 + 0.50
Commands
llama.cpp
llama-server -hf unsloth/Qwen3.5-4B-GGUF:IQ4_NL
Ollama
ollama run hf.co/unsloth/Qwen3.5-4B-GGUF:IQ4_NL
Download
hf download unsloth/Qwen3.5-4B-GGUF --include "Qwen3.5-4B-IQ4_NL.gguf" --local-dir Qwen3.5-4B-GGUF
Q4_02.4 GBfile size3.2 GB2.4 + 0.25 + 0.50
Commands
llama.cpp
llama-server -hf unsloth/Qwen3.5-4B-GGUF:Q4_0
Ollama
ollama run hf.co/unsloth/Qwen3.5-4B-GGUF:Q4_0
Download
hf download unsloth/Qwen3.5-4B-GGUF --include "Qwen3.5-4B-Q4_0.gguf" --local-dir Qwen3.5-4B-GGUF
Q4_K_S2.4 GBfile size3.2 GB2.4 + 0.25 + 0.50
Commands
llama.cpp
llama-server -hf unsloth/Qwen3.5-4B-GGUF:Q4_K_S
Ollama
ollama run hf.co/unsloth/Qwen3.5-4B-GGUF:Q4_K_S
Download
hf download unsloth/Qwen3.5-4B-GGUF --include "Qwen3.5-4B-Q4_K_S.gguf" --local-dir Qwen3.5-4B-GGUF
Q4_K_M2.6 GBfile size3.3 GB2.6 + 0.25 + 0.50
Commands
llama.cpp
llama-server -hf unsloth/Qwen3.5-4B-GGUF:Q4_K_M
Ollama
ollama run hf.co/unsloth/Qwen3.5-4B-GGUF:Q4_K_M
Download
hf download unsloth/Qwen3.5-4B-GGUF --include "Qwen3.5-4B-Q4_K_M.gguf" --local-dir Qwen3.5-4B-GGUF
Q4_12.6 GBfile size3.3 GB2.6 + 0.25 + 0.50
Commands
llama.cpp
llama-server -hf unsloth/Qwen3.5-4B-GGUF:Q4_1
Ollama
ollama run hf.co/unsloth/Qwen3.5-4B-GGUF:Q4_1
Download
hf download unsloth/Qwen3.5-4B-GGUF --include "Qwen3.5-4B-Q4_1.gguf" --local-dir Qwen3.5-4B-GGUF
UD-Q4_K_XL2.7 GBfile size3.5 GB2.7 + 0.25 + 0.50
Commands
llama.cpp
llama-server -hf unsloth/Qwen3.5-4B-GGUF:UD-Q4_K_XL
Ollama
ollama run hf.co/unsloth/Qwen3.5-4B-GGUF:UD-Q4_K_XL
Download
hf download unsloth/Qwen3.5-4B-GGUF --include "Qwen3.5-4B-UD-Q4_K_XL.gguf" --local-dir Qwen3.5-4B-GGUF
Q5_K_S2.8 GBfile size3.6 GB2.8 + 0.25 + 0.50
Commands
llama.cpp
llama-server -hf unsloth/Qwen3.5-4B-GGUF:Q5_K_S
Ollama
ollama run hf.co/unsloth/Qwen3.5-4B-GGUF:Q5_K_S
Download
hf download unsloth/Qwen3.5-4B-GGUF --include "Qwen3.5-4B-Q5_K_S.gguf" --local-dir Qwen3.5-4B-GGUF
Q5_K_M2.9 GBfile size3.7 GB2.9 + 0.25 + 0.50
Commands
llama.cpp
llama-server -hf unsloth/Qwen3.5-4B-GGUF:Q5_K_M
Ollama
ollama run hf.co/unsloth/Qwen3.5-4B-GGUF:Q5_K_M
Download
hf download unsloth/Qwen3.5-4B-GGUF --include "Qwen3.5-4B-Q5_K_M.gguf" --local-dir Qwen3.5-4B-GGUF
UD-Q5_K_XL3.0 GBfile size3.8 GB3.0 + 0.25 + 0.50
Commands
llama.cpp
llama-server -hf unsloth/Qwen3.5-4B-GGUF:UD-Q5_K_XL
Ollama
ollama run hf.co/unsloth/Qwen3.5-4B-GGUF:UD-Q5_K_XL
Download
hf download unsloth/Qwen3.5-4B-GGUF --include "Qwen3.5-4B-UD-Q5_K_XL.gguf" --local-dir Qwen3.5-4B-GGUF
Q6_K3.3 GBfile size4.0 GB3.3 + 0.25 + 0.50
Commands
llama.cpp
llama-server -hf unsloth/Qwen3.5-4B-GGUF:Q6_K
Ollama
ollama run hf.co/unsloth/Qwen3.5-4B-GGUF:Q6_K
Download
hf download unsloth/Qwen3.5-4B-GGUF --include "Qwen3.5-4B-Q6_K.gguf" --local-dir Qwen3.5-4B-GGUF
UD-Q6_K_XL3.9 GBfile size4.6 GB3.9 + 0.25 + 0.50
Commands
llama.cpp
llama-server -hf unsloth/Qwen3.5-4B-GGUF:UD-Q6_K_XL
Ollama
ollama run hf.co/unsloth/Qwen3.5-4B-GGUF:UD-Q6_K_XL
Download
hf download unsloth/Qwen3.5-4B-GGUF --include "Qwen3.5-4B-UD-Q6_K_XL.gguf" --local-dir Qwen3.5-4B-GGUF
Q8_04.2 GBfile size4.9 GB4.2 + 0.25 + 0.50
Commands
llama.cpp
llama-server -hf unsloth/Qwen3.5-4B-GGUF:Q8_0
Ollama
ollama run hf.co/unsloth/Qwen3.5-4B-GGUF:Q8_0
Download
hf download unsloth/Qwen3.5-4B-GGUF --include "Qwen3.5-4B-Q8_0.gguf" --local-dir Qwen3.5-4B-GGUF
UD-Q8_K_XL5.5 GBfile size6.3 GB5.5 + 0.25 + 0.50
Commands
llama.cpp
llama-server -hf unsloth/Qwen3.5-4B-GGUF:UD-Q8_K_XL
Ollama
ollama run hf.co/unsloth/Qwen3.5-4B-GGUF:UD-Q8_K_XL
Download
hf download unsloth/Qwen3.5-4B-GGUF --include "Qwen3.5-4B-UD-Q8_K_XL.gguf" --local-dir Qwen3.5-4B-GGUF
BF167.8 GBfile size8.6 GB7.8 + 0.25 + 0.50
Commands
llama.cpp
llama-server -hf unsloth/Qwen3.5-4B-GGUF:BF16
Ollama
ollama run hf.co/unsloth/Qwen3.5-4B-GGUF:BF16
Download
hf download unsloth/Qwen3.5-4B-GGUF --include "Qwen3.5-4B-BF16.gguf" --local-dir Qwen3.5-4B-GGUF

MLX for Apple silicon

FormatSizeMemory neededRepo
MLX 4bitmlx-community/Qwen3.5-4B-MLX-4bit2.8 GBfile size3.6 GB2.8 + 0.25 + 0.50Download page
MLXnace-ai/Qwen3.5-4B-MLX5.7 GBfile size6.4 GB5.7 + 0.25 + 0.50Download page

Original weights size from the safetensors header, for vLLM, SGLang, Transformers

FormatSizeMemory neededRepo
BF16original weights8.7 GBfile size9.4 GB8.7 + 0.25 + 0.50
hf download Qwen/Qwen3.5-4B --local-dir Qwen3.5-4B
Other conversions (12; sizes not read)

The Hub tags 100 or more quantised repos as conversions of this model; we read the most downloaded and show the maker's own first.

Run it by API

Hugging Face Inference Providers, fetched 19 Nov 2026, 13:58 UTC

No provider on Hugging Face Inference Providers serves this model (as of 19 Nov 2026, 13:58 UTC). Other hosts may; we only list what the router reports.

Not built yet: one key for every open model, with BlockRun, and our own tested-quality score per provider. Until then, precision and measured speed are the best guide to a provider here.

Fine-tunes of this model

from our snapshot of the Hub

Repos that declare this model as their base, most downloaded first. Listing is not a recommendation; unreviewed ones have not been checked.

Planned for this page Not built yet

Try it in private chat
A link to Solana AI's confidential chat for the models it serves. This model is not listed as served there.
Fine-tune it
Fine-tuning a few base models from this page and publishing the result here.
Creators and earnings
Pages for people who publish fine-tunes and adapters of this model, and what they earn.
Fund and bounties
Bounties for new open models, paid from a public fund. How the fund will work.
Contribute compute
How to give compute to open model work, with Akash.

Model data from the Hugging Face Hub 19 Nov 2026, 13:43 UTC. The repo. Found a mistake on this page? Tell us.