[whatmodel.sh](../../README.md) / Qwen3.5 0.8B # Qwen3.5 0.8B Alibaba · released 2026-03 · 0.87B params · apache-2.0 **score 37.7** ±0.9 · #600 of 732 · from 56 benchmark results across 8 sources Score within the [LMArena (text)](https://whatmodel.sh/?set=lmarena-text&score=set) comparison set only. ## What it's good at _{● marks Qwen3.5 0.8B among every model on that task or test, worst to best.}

task                vs the others          rank  counts  tests
agent tasks         ▁●▃▆█▄▅▅▄▄▂▁   #304 of 313*    3.4%      2
chat & writing      ▁▁▁▃●▄▆█▆▇▇▃   #357 of 402*    1.3%      1
math                ▁▁▃●█▆▄█▄▂▁▁    #382 of 431    5.9%      3
vision & documents  ▁▁●▂▁▃▄▂▆▇█▄    #140 of 153     25%     18

agent tasks · #304 of 313 · 3.4% of its score

Beats [QwQ-32B-Preview](qwq-32b-preview.md), [DeepSeek-Coder-V2 Lite Instruct (16B-A2.4B)](deepseek-coder-v2-lite-instruct.md), [Phi-4-mini 3.8B](phi-4-mini.md) here, though they rank higher overall. Trails [Granite 4.0 350M](granite-4-0-350m.md), [Granite 4.0 H 350M](granite-4-0-h-350m.md), [Qwen2.5 0.5B Instruct](qwen2-5-0-5b.md) here, though they rank lower overall.

test                                result  vs the others in its run   difficulty        counts  why
τ²-bench aggregate (maker)           11.6%  ●▁▁·▁▁▃▃▃▄█▁ beats 0/39    medium · avg 46%    1.9%  small run
BFCL V4 (December 2025 archive) (…   25.3%  ▁▁▁▂●▆▅▆█▂▄▃ beats 16/114  medium · avg 43%    1.6%  small run

chat & writing · #357 of 402 · 1.3% of its score

Beats [Code Llama 70B Instruct](codellama-70b-instruct.md), [Gemma 4 E2B](gemma-4-e2b.md), [Phi-3-mini-4k-instruct (June 2024 update)](phi-3-mini-4k-instruct-june-2024.md) here, though they rank higher overall. Trails [Gemma 3 270M IT](gemma-3-270m.md), [SmolLM2 360M Instruct](smollm2-360m.md), [Granite 4.0 H 350M](granite-4-0-h-350m.md) here, though they rank lower overall.

test                                result  vs the others in its run  difficulty      counts  why
IFEval (aggregation unspecified) …   44.0%  ▁●▄▂▂▄▆██▅▂▁ beats 4/145  easy · avg 76%    1.3%  at its level · small run

math · #382 of 431 · 5.9% of its score

Beats [Yi 34B Chat](yi-34b-chat.md), [Llama 3 8B Instruct](llama-3-8b.md), [Command R (03-2024)](command-r-03-2024.md) here, though they rank higher overall. Trails [GPT-Neo 125M](gpt-neo-125m.md), [GPT-Neo 1.3B](gpt-neo-1-3b.md), [OPT-1.3B](opt-1-3b.md) here, though they rank lower overall.

test                          result  vs the others in its run  difficulty      counts  why
MathVista (testmini) (maker)   62.2%  ▁··▁▄▄▆●▄▅█▄ beats 17/48  easy · avg 75%    2.4%  at its level · small run
DynaMath (maker)               49.9%  ●·····▂··█▄▆ beats 0/10   easy · avg 71%    1.7%  at its level · small run
PolyMATH (maker)                8.2%  ●···▃▁·▇█▄▄▄ beats 0/23   hard · avg 29%    1.7%  small run

vision & documents · #140 of 153 · 25% of its score

Beats [Gemma 3 4B](gemma-3-4b.md), [LLaVA-OneVision Qwen2 72B (ov-sft)](llava-onevision-qwen2-72b-ov.md), [CogVLM2 Llama3 Chat 19B](cogvlm2-llama3-chat-19b.md) here, though they rank higher overall.

test                                result  vs the others in its run   difficulty           counts  why
CharXiv (reasoning questions) (ma…   41.3%  ▂▅▃●▅▆▆▅▆█▆▂ beats 20/101  medium · avg 58%       2.6%  at its level · small run
MathVista (testmini) (maker)         62.2%  ▁··▁▄▄▆●▄▅█▄ beats 17/48   easy · avg 75%         2.4%  at its level · small run
ChartArena chart parsing (English…    9.8%  ●·▁·▃▃▁▇▃█·▄ beats 0/22    hard · avg 27%         2.4%  blunt
MMMU (val) (maker)                   49.0%  ▁····▁▁●█▆▆▅ beats 8/53    medium · avg 64%       2.2%  at its level · small run
MMMU Pro (maker)                     31.2%  ●▃▁▃▆▇▃▇▇██▅ beats 0/52    medium · avg 50%       1.9%  small run
OmniDocBench 1.5 overall (higher …   61.0%  ●··▂▂▂··▂▂█▅ beats 0/13    easy · avg 76%         1.8%  at its level · small run
MMStar (maker)                       58.3%  ▃▅▂●▂▂▆██▆█▅ beats 7/35    medium · avg 70%       1.7%  at its level · small run
CountBench (maker)                   77.0%  ▂▂▃●·▂▂▃▃·▅█ beats 4/18    very easy · avg 90%    1.5%  small run
PMC-VQA (maker)                      40.4%  ●···▃··▃▃·▃█ beats 0/7     medium · avg 50%       1.2%  at its level · small run
MMBench-EN dev v1.1 (maker)          69.9%  ●▄···▂··▄▂▄▆ beats 3/14    easy · avg 82%         1.2%  blunt · at its level · small run
SimpleVQA (maker)                    31.3%  ●·▃·▃·▃·█▅▃▃ beats 0/10    medium · avg 41%       1.2%  blunt · small run
BEAR-Bench (bilingual RU/EN enter…   10.6%  ●▃·▅·▃·█·█▅▅ beats 0/14    hard · avg 24%         1.1%  typical
HallusionBench (maker)               53.1%  ▂▁·▂●▂▃▅▃█▅▅ beats 6/33    medium · avg 59%       0.8%  blunt · at its level · small run
RealWorldQA (maker)                  63.4%  ▁▃▁●▃▇▇▅█▁▅▅ beats 6/39    easy · avg 73%         0.7%  blunt · at its level · small run
CC-OCR (maker)                       63.2%  ●·▂··▂▂▃▃▂▆█ beats 0/26    easy · avg 73%         0.6%  blunt · at its level · small run
OCRBench (maker)                     74.5%  ▁·●▁▆▅▂█▅█▁▃ beats 3/37    easy · avg 83%         0.6%  blunt · small run
AI2D test (maker)                    69.9%  ▁▁●▇▃▇·▄·█·▃ beats 2/28    easy · avg 84%         0.3%  blunt · at its level · small run
PureDocBench document parsing (Av…   56.0%  ▂·●·▂▂▅█·▁▂▂ beats 2/22    medium · avg 62%       0.2%  blunt · at its level

other tests (34)

test                                result  vs the others in its run   difficulty           counts  why
EuroEval Life in the UK (English …   56.3%  ●·▂▂·▂·▃▃·█▃ beats 0/14    easy · avg 80%         7.9%  at its level
MMLU-Pro (maker)                     42.3%  ▁▁·▁▁▁●▂▄▅█▁ beats 32/188  medium · avg 66%       4.5%  at its level
MMLU-Redux (maker)                   59.5%  ●▁▃▁▂▃▄▅▇▅█▄ beats 0/63    easy · avg 81%         3.6%  at its level · small run
C-Eval (maker)                       50.5%  ▁···▁▁▂●▂▂▃█ beats 7/33    easy · avg 75%         3.5%  at its level · small run
INCLUDE (maker)                      40.6%  ▂····●▂▄▄▄█▄ beats 2/36    medium · avg 59%       3.2%  small run
MMMLU (maker)                        44.3%  ▂▁▁▂●▂▄▄▅██▃ beats 7/46    medium · avg 65%       3.0%  at its level · small run
EuroEval HellaSwag (English), val…   38.0%  ●··▂▂·▂▆▆█·▄ beats 0/15    medium · avg 63%       2.7%  at its level
MAXIFE (maker)                       39.2%  ●·▂·▂···▆·▄█ beats 0/11    medium · avg 63%       2.3%  at its level · small run
MMLU-ProX (maker)                    34.6%  ●·▁▁▁▂▁▃▃█▅▄ beats 0/36    medium · avg 57%       2.2%  at its level · small run
MultiChallenge (maker)               18.9%  ●·▁·▁▃▆▅▃▂█▅ beats 0/30    hard · avg 35%         2.1%  small run
SuperGPQA (maker)                    21.3%  ●▁▂▂▅▆▃▇▇▇█▃ beats 0/46    medium · avg 40%       2.1%  small run
Global PIQA (maker)                  59.4%  ●▃··▃··▃▃▅█▅ beats 0/11    medium · avg 70%       2.0%  small run
IFBench (aggregation unspecified)…   21.0%  ▂●▂▄▄▄▂▄▄▆█▇ beats 8/86    medium · avg 40%       2.0%  small run
KINA (Knowledge Index of Noah's A…   16.7%  ▁●▂▁▃▄█▆▄▅▂▃ beats 1/39    very hard · avg 21%    1.8%  too hard for it
VLMs Are Blind (maker)               59.4%  ●·▂·····▃··█ beats 0/8     easy · avg 82%         1.7%  at its level · small run
MedXPertQA MM (maker)                17.1%  ●···▂·▃··▆▂█ beats 0/13    very hard · avg 28%    1.7%  sharp · too hard for it · small run
VideoMMMU (maker)                    44.3%  ●···▂···▂▂█▆ beats 0/10    medium · avg 63%       1.7%  at its level · small run
Video-MME (with subtitles) (maker)   63.8%  ▁▁▁●▁·▁▃▁▁▁█ beats 3/16    easy · avg 75%         1.6%  at its level · small run
SLAKE (maker)                        62.6%  ●·····▄▄·██▄ beats 0/7     easy · avg 73%         1.6%  at its level · small run
MLVU (maker)                         65.6%  ●···▂···▂▄▄█ beats 0/10    easy · avg 76%         1.5%  at its level · small run
LVBench (maker)                      45.1%  ●·█▅▅▄▄··▃▁▁ beats 0/28    medium · avg 56%       1.5%  small run
MMVU (maker)                         34.3%  ●···▃···▃▃▅█ beats 0/8     medium · avg 52%       1.4%  at its level · small run
EmbSpatial-Bench (maker)             68.6%  ●·····▁·▁·▃█ beats 0/10    easy · avg 76%         1.4%  at its level · small run
ERQA (maker)                         34.5%  ●··▃··█·▃█▅▅ beats 0/12    medium · avg 45%       1.4%  at its level · small run
RefSpatial-Bench (maker)             23.5%  ●·▃····▃▃▅▅█ beats 0/10    medium · avg 40%       1.3%  small run
MMLongBench-Doc (maker)              33.6%  ●···▃···▃▅█▅ beats 0/9     medium · avg 45%       1.3%  at its level · small run
NOVA-63 (maker)                      42.4%  ▃●·▃···▃█▃▃█ beats 1/11    medium · avg 47%       1.3%  blunt · at its level · small run
Video-MME (without subtitles) (ma…   57.7%  ▁▃●▃▁·█▄▄▅·▇ beats 3/27    medium · avg 69%       1.2%  blunt · at its level · small run
ZEROBench subquestions (maker)       12.9%  ●·▄···▄·▄██▄ beats 0/8     hard · avg 19%         1.2%  small run
LongBench v2 (maker)                 26.1%  ●·▁·▂▃▁▄▄▃█▄ beats 0/33    very hard · avg 34%    0.9%  too hard for it · noisy · small run
MVBench (maker)                      55.8%  ●··▃▄▁▃▄▄▄█▄ beats 0/26    medium · avg 66%       0.6%  blunt · at its level · small run
LIBRA Mini (Russian long context,…   35.5%  ▄···●█·██··▄ beats 1/8     medium · avg 38%       0.4%  blunt · at its level
WMT24++ (maker)                      27.2%  ●·▁▁·▁·▁▂▆█▃ beats 0/27    medium · avg 62%       0.3%  blunt · at its level · small run
GPQA (subset unspecified) (maker)    11.9%  ▂▁▂▂▄▃▆▃▅█▃▃ beats 0/106   very hard · avg 32%       —  failed run: 11.9% on GPQA (4 options, c…

## How the score was worked out Each result implies a score on its own, after adjusting for how hard that test is and who ran it. The model's score is their weighted average, with a small pull toward the middle for thinly tested models. Each test's weight is its *counts* column above; here it is added up by who ran the tests. [Method](../method.md)

source                                    results  share of the score
Qwen/Qwen3.5-2B model card (submitted /…       48   79% ▇▇▇▇▇▇▇▇▇
EuroEval leaderboards                           2   11% ▇
Qwen/Qwen3.5-0.8B model card (submitted…        1    5% ▇
ChartArena: Benchmarking Chart Parsing …        1    2% ▇
Knowledge Index of Noah's Ark                   1    2% ▇
BEAR-Bench: A Bilingual Enterprise and …        1    1% ▇
ai-forever LIBRA leaderboard                    1    0% ▇
How Far Is Document Parsing from Solved…        1    0% ▇
prior: pull toward the middle                        0% ▇
weighted average                                   = 37.7

_{Licences: Qwen/Qwen3.5-2B model card ([apache-2.0](../../licenses/Qwen--Qwen3.5-2B.txt)) · EuroEval leaderboards ([MIT](https://github.com/EuroEval/leaderboards/blob/1a77f1d2f8ba1ba05ab73287fb090d76bf41b2e0/LICENSE)) · Qwen/Qwen3.5-0.8B model card ([apache-2.0](../../licenses/Qwen--Qwen3.5-0.8B.txt)) · ChartArena: Benchmarking Chart Parsing across Languages ([CC BY 4.0](https://creativecommons.org/licenses/by/4.0/)) · Knowledge Index of Noah's Ark ([CC BY 4.0](https://creativecommons.org/licenses/by/4.0/)) · BEAR-Bench: A Bilingual Enterprise and Academic Reasoning Benchmark for Multimodal Models ([CC BY 4.0](https://creativecommons.org/licenses/by/4.0/)) · ai-forever LIBRA leaderboard ([MIT](../../licenses/hf-leaderboard-spaces-MIT.txt)) · How Far Is Document Parsing from Solved? PureDocBench: A Source-Traceable Benchmark across Clean ([CC BY 4.0](https://creativecommons.org/licenses/by/4.0/))}

Not counted (1)

source                  test                       score  why
Qwen/Qwen3.5-2B model…  GPQA (subset unspecified)  11.9%  failed run: 11.9% on GPQA (4 options, chance 25%) in thinking mode; probably unterminated thinking or unparsed answers rather than below-random knowledge. The next-lowest model on this benchmark scores 25.2%.

## How to run it ### llama.cpp (recommended) 1. Install llama.cpp. Homebrew and winget installs name the commands llama-cli and llama-server instead. [Other ways to install](https://github.com/ggml-org/llama.cpp/blob/master/docs/install.md) ```sh # macOS / Linux curl -LsSf https://llama.app/install.sh | sh # Windows irm https://llama.app/install.ps1 | iex ``` 2. Run Qwen3.5 0.8B (Q4_K_M, needs ~0.5 GB, from [huggingface.co/lmstudio-community](https://huggingface.co/lmstudio-community/Qwen3.5-0.8B-GGUF/tree/26bab2c9369648924251c0ebb3dae012f5147707)) ```sh llama cli -hf lmstudio-community/Qwen3.5-0.8B-GGUF:Q4_K_M ``` 3. Or start a local server and open the address it prints (http://localhost:9931) ```sh llama serve -hf lmstudio-community/Qwen3.5-0.8B-GGUF:Q4_K_M ``` Other versions: Q8_0 (~0.8 GB) `llama cli -hf lmstudio-community/Qwen3.5-0.8B-GGUF:Q8_0` ### LM Studio 1. Install LM Studio. Open the app once before using the lms command. [Download it (macOS / Windows / Linux)](https://lmstudio.ai/download) 2. Run Qwen3.5 0.8B (Q4_K_M, needs ~0.5 GB, from [huggingface.co/lmstudio-community](https://huggingface.co/lmstudio-community/Qwen3.5-0.8B-GGUF/tree/26bab2c9369648924251c0ebb3dae012f5147707)) ```sh lms get https://huggingface.co/lmstudio-community/Qwen3.5-0.8B-GGUF@Q4_K_M -y ``` 3. Then run lms ls to see its name, and lms chat followed by that name Other versions: Q8_0 (~0.8 GB) `lms get https://huggingface.co/lmstudio-community/Qwen3.5-0.8B-GGUF@Q8_0 -y` · 4bit (~0.6 GB, Mac only) `lms get https://huggingface.co/lmstudio-community/Qwen3.5-0.8B-MLX-4bit -y` · 8bit (~1 GB, Mac only) `lms get https://huggingface.co/lmstudio-community/Qwen3.5-0.8B-MLX-8bit -y` ### Ollama 1. Install Ollama. [Download it (macOS / Windows)](https://ollama.com/download) ```sh # Linux curl -fsSL https://ollama.com/install.sh | sh ``` 2. Run Qwen3.5 0.8B (Q8_0, needs ~1.54 GB, from [ollama.com](https://ollama.com/library/qwen3.5/tags)) ```sh ollama run qwen3.5:0.8b ``` Other versions: mxfp8 (~1.02 GB, Mac only) `ollama run qwen3.5:0.8b-mlx` ### MLX (Mac) 1. Install MLX (Mac). Apple silicon Macs only, text only. ```sh # with uv uv tool install mlx-lm # or pip install mlx-lm ``` 2. Run Qwen3.5 0.8B (4bit, needs ~0.6 GB, Mac only, from [huggingface.co/lmstudio-community](https://huggingface.co/lmstudio-community/Qwen3.5-0.8B-MLX-4bit/tree/c9892b0295bceb223ffd3229aebd5a7ac73bb849)) ```sh mlx_lm.chat --model lmstudio-community/Qwen3.5-0.8B-MLX-4bit ``` 3. Or start an OpenAI-compatible server at http://localhost:8080/v1 ```sh mlx_lm.server --model lmstudio-community/Qwen3.5-0.8B-MLX-4bit ``` Other versions: 8bit (~1 GB, Mac only) `mlx_lm.chat --model lmstudio-community/Qwen3.5-0.8B-MLX-8bit` Model repository and documentation: [huggingface.co/Qwen](https://huggingface.co/Qwen/Qwen3.5-0.8B) ## Published speed measurements Reference results for the hardware, engine and settings below. These are not verified speeds of the downloads above or predictions for your machine.

hardware                      precision               speed  engine and settings
Apple M5 Pro (64GB; CPU/GPU…  BF16 (paper's m…  113.4 tok/s  omlx (MLX; SSD prefix cache, in-m…
Apple M5 Pro (64GB; CPU/GPU…  BF16 (paper's m…  105.5 tok/s  llama.cpp server (Metal; --parall…
Apple M5 Pro (64GB; CPU/GPU…  BF16 (paper's m…  100.9 tok/s  vllm-metal (Metal; version not st…
Apple M5 Pro (64GB; CPU/GPU…  BF16 (paper's m…   95.3 tok/s  vllm-mlx (MLX; version not stated…
Apple M5 Pro (64GB; CPU/GPU…  BF16 (paper's m…   93.9 tok/s  mlx_lm server (MLX; version not s…
Apple M5 Pro (64GB; CPU/GPU…  BF16 (paper's m…     77 tok/s  omlx (MLX; SSD prefix cache, in-m…
Apple M5 Pro (64GB; CPU/GPU…  BF16 (paper's m…   76.2 tok/s  vllm-metal (Metal; version not st…
Apple M5 Pro (64GB; CPU/GPU…  BF16 (paper's m…   67.6 tok/s  vllm-mlx (MLX; version not stated…
Apple M5 Pro (64GB; CPU/GPU…  BF16 (paper's m…   66.9 tok/s  llama.cpp server (Metal; --parall…
NVIDIA DGX Spark (GB10)       BF16              106.5 tok/s  llama.cpp (CUDA build, GGUF BF16 …
NVIDIA DGX Spark (GB10)       BF16               77.5 tok/s  llama.cpp (CUDA build, GGUF BF16 …
NVIDIA DGX Spark (GB10)       BF16               94.2 tok/s  vLLM (CUDA, Safetensors BF16; ver…
NVIDIA DGX Spark (GB10)       BF16               73.4 tok/s  vLLM (CUDA, Safetensors BF16; ver…
NVIDIA DGX Spark (GB10)       BF16                 95 tok/s  SGLang (CUDA, Safetensors BF16; v…
NVIDIA DGX Spark (GB10)       BF16               79.4 tok/s  SGLang (CUDA, Safetensors BF16; v…
Apple M5 Max (40-core GPU),…  mlx-mxfp8         319.3 tok/s  ollama (version not recorded) · l…

_{Each speed links to its source, which has the full settings.} [interactive version](https://whatmodel.sh/model/qwen3-5-0-8b) · [JSON](https://whatmodel.sh/api?format=json&all=1)