[whatmodel.sh](../../README.md) / Qwen3 32B
# Qwen3 32B
Alibaba · released 2025-04 · 32.8B params · apache-2.0
**score 57.4** ±0.4 · #237 of 732 · from 103 benchmark results across 23 sources
Score within the [LMArena (text)](https://whatmodel.sh/?set=lmarena-text&score=set) comparison set only.
## What it's good at
● marks Qwen3 32B among every model on that task or test, worst to best.
task vs the others rank counts tests
agent tasks ▁▁▃▆●▄▅▅▄▄▂▁ #196 of 313 22% 10
coding ▁▁▁▃▅▇█●▅▄▃▁ #205 of 469 13% 8
chat & writing ▁▁▁▃▆▄▆█●▇▇▃ #171 of 402 7.8% 5
math ▁▁▃▅█▆●█▄▂▁▁ #173 of 431 17% 8
agent tasks · #196 of 313 · 22% of its score
Beats [o1 (2024-12-17)](o1-2024-12-17.md), [o3-mini](o3-mini.md), [DeepSeek R1 0528](deepseek-r1-0528.md) here, though they rank higher overall. Trails [SmolLM3 3B](smollm3-3b.md), [Jamba 1.5 Mini](jamba-1-5-mini.md), [Mellum2 12B-A2.5B Instruct](mellum2-12b-a2-5b-instruct.md) here, though they rank lower overall.
test result vs the others in its run difficulty counts why
τ²-bench Telecom (maker) 26.3% ▃·●········█ beats 2/6 very hard · avg 9.9% 10% sharp · several runs
τ-bench Retail (maker) 52.8% ▁····▃·●▇█▂▆ beats 5/28 medium · avg 37% 2.8% at its level · several runs
τ-bench Airline (maker) 29.0% ▁▁▁·▅●▁█▇▁▁▇ beats 9/28 hard · avg 25% 1.8% at its level · small run · several runs
τ²-bench Retail (maker) 49.7% ▁····▂●▃▆█▃▁ beats 5/44 medium · avg 37% 1.8% at its level · several runs
τ²-bench Airline (maker) 45.5% ▁▁·▂▁●▂▁█▅▂▃ beats 8/42 hard · avg 32% 1.3% at its level · several runs
Terminal-Bench Hard 2.8% ▂●▆▂▂▆▂▂█▅▅▂ beats 3/27 very hard · avg 1.3% 1.3% sharp · too hard for it
BFCL v3 69.3% ▁▁▁▁▁▃▅█▅▄●▁ beats 108/114 medium · avg 51% 1.0% blunt · at its level · several runs
SWE-bench Verified 23.4% ▁▁▁▁▁●▂▄█▆▁▁ beats 14/138 hard · avg 16% 0.9% sharp · at its level · small run
BFCL V4 (December 2025 archive) 48.7% ▁▁▁▂██▅●▆▂▂▂ beats 56/77 hard · avg 31% 0.8% at its level
Multi-SWE-bench (scaffold per not… 7.7% ●··█·······█ beats 0/2 very hard · avg 5.9% 0.3% small run
coding · #205 of 469 · 13% of its score
Beats [Gemini 2.5 Flash](gemini-2-5-flash.md), [Qwen3 30B-A3B Instruct 2507](qwen3-30b-a3b-instruct-2507.md), [Qwen3.5 4B](qwen3-5-4b.md) here, though they rank higher overall. Trails [QwQ-32B-Preview](qwq-32b-preview.md), [Ministral 3 3B Reasoning 2512](ministral-3-3b-reasoning.md), [DeepSeek-Coder-V2 Instruct (0724)](deepseek-coder-v2-0724.md) here, though they rank lower overall.
test result vs the others in its run difficulty counts why
Aider polyglot (225 exercises, tw… 40.0% ▂·▂▃▅▂●█▆▄▂▂ beats 20/49 very hard · avg 7.6% 6.5% sharp · at its level · several runs
LiveCodeBench v6 (Feb–May 2025) 53.4% ▁▁▃▁█●▄▇▅▅▃█ beats 14/38 medium · avg 41% 2.0% at its level · several runs
LiveCodeBench v6 (window unspecif… 61.0% █·▄·●····▄·▄ beats 3/5 medium · avg 41% 1.2% at its level
WebApp1K-Duo-React pass@1 (onekq-… 24.3% ▁··●·▁▁▁▆█▇█ beats 2/45 hard · avg 29% 1.0% at its level
SWE-bench Verified 23.4% ▁▁▁▁▁●▂▄█▆▁▁ beats 14/138 hard · avg 16% 0.9% sharp · at its level · small run
MultiPL-E (maker) 76.9% ▃··▄·▅▅▇▄●▃▄ beats 24/31 easy · avg 68% 0.7% typical
Multi-SWE-bench (scaffold per not… 7.7% ●··█·······█ beats 0/2 very hard · avg 5.9% 0.3% small run
SciCode main problems 9.2% ▄··▄····█▄·● beats 5/5 very hard · avg 3.8% 0.3% too hard for it
chat & writing · #171 of 402 · 7.8% of its score
Beats [MiniMax-M2](minimax-m2.md), [MiniCPM5 2B](minicpm5-2b.md), [Granite 4.2 30B](granite-4-2-30b.md) here, though they rank higher overall. Trails [LFM2.5 1.2B Instruct](lfm2-5-1-2b-instruct.md), [LFM2.5 1.2B Thinking](lfm2-5-1-2b-thinking.md), [LFM2.5 8B-A1B](lfm2-5-8b-a1b.md) here, though they rank lower overall.
test result vs the others in its run difficulty counts why
Arena-Hard v2 aggregate (maker) 48.4% ▁▁··▄▃●▃█▁▁▇ beats 7/22 hard · avg 28% 2.4% sharp · at its level · several runs
Arena-Hard v2.0 hard prompts (sty… 44.5% ▂·▃▅▅·▃●▆▂▂▂ beats 12/22 very hard · avg 12% 1.8% sharp · at its level
LMArena (text) 1346.8 ▁▁▁▂▃▃▃▆●▇█▃ beats 189/385 avg 1260 1.7% blunt
IFEval (aggregation unspecified) … 85.0% ▁▂▄▂▂▄▆█●▅▂▁ beats 99/145 easy · avg 76% 1.4% several runs
Arena-Hard v2.0 creative writing … 53.3% ▆▂▂▆▂▂●█▂▃·▂ beats 13/22 hard · avg 19% 0.4% sharp · at its level · noisy
math · #173 of 431 · 17% of its score
Beats [MiMo-V2.5 (310B-A15B)](mimo-v2-5.md), [MiniMax-M2.1](minimax-m2-1.md), [Mistral Medium 3.5 (128B)](mistral-medium-3-5.md) here, though they rank higher overall. Trails [OpenThinker3 7B](openthinker3-7b.md), [Olmo 3 7B Think](olmo-3-7b-think.md), [OpenThinker2 32B](openthinker2-32b.md) here, though they rank lower overall.
test result vs the others in its run difficulty counts why
AIME 2025 71.0% ▄·●··▄····▄▄ beats 1/5 medium · avg 52% 5.2% sharp · at its level · several runs
HMMT 2025 (edition unspecified) 49.8% ▄··▄●·▄····█ beats 2/5 hard · avg 29% 4.8% sharp · at its level · several runs
AIME 2024 79.4% █●···█·····▄ beats 2/5 medium · avg 62% 3.5% sharp · several runs
OTIS mock AIME 2024–2025 66.9% ▃▃▄▃▃▃●▇█▃▂▅ beats 96/197 very hard · avg 7.2% 1.2% sharp · at its level · noisy
PolyMATH (maker) 47.4% ▁···▃▁·▇●▄▄▄ beats 9/23 hard · avg 32% 1.2% at its level · several runs
MATH-500 96.8% ▄·▄·····●··█ beats 2/5 very easy · avg 92% 0.8% maxed out
OMEGA (math generalisation) 47.7% ▅···▃▃···▃●█ beats 5/8 hard · avg 24% 0.4% at its level
MATH (subset unspecified) 95.4% ▅····▃▃··▃●█ beats 5/8 easy · avg 73% 0.2% noisy
other tests (34)
test result vs the others in its run difficulty counts why
GPQA Diamond 65.7% ▁▁▁▁▂▄▆▄●▇█▅ beats 97/211 hard · avg 39% 7.5% sharp · at its level · several runs
MMLU-Pro 73.9% █··█··█●█··█ beats 3/5 medium · avg 58% 4.0% at its level · several runs
GPQA (subset unspecified) (maker) 68.4% ▂▁▂▂▄▃▆▃▅●▃▃ beats 74/106 hard · avg 49% 3.7% sharp · at its level · several runs
MMLU-Redux (maker) 90.9% ▁▁▃▁▂▃▄▅▇●█▄ beats 40/63 easy · avg 78% 2.1% several runs
INCLUDE (maker) 73.7% ▂····▂▂▄▄●█▄ beats 20/36 medium · avg 57% 2.0% at its level · several runs
Humanity's Last Exam 8.3% █··██···●·██ beats 3/5 very hard · avg 3.7% 1.8% too hard for it · several runs
SuperGPQA (maker) 54.1% ▂▁▂▂▅▆▃▇●▇█▃ beats 25/46 medium · avg 39% 1.8% at its level · several runs
MMLU 86.2% ▁▁▁▁▁▂▂▅██▆● beats 194/214 medium · avg 70% 1.4% typical
MMLU-ProX (maker) 74.6% ▁·▁▁▁▂▁▃●█▅▄ beats 12/36 medium · avg 61% 1.3% at its level · several runs
WritingBench (maker) 79.0% ▁·▁●▁··▅█▂▂▁ beats 3/20 easy · avg 71% 1.1% several runs
KCL-Essay vanilla 29.2% ▁···▁▃▂●▂▃▃▂ beats 7/23 hard · avg 17% 1.1% at its level
BRIDGE clinical text, zero-shot a… 41.0% ▁▁▁▁▁▂▂▃▄▆●▃ beats 82/97 hard · avg 31% 1.1% blunt · at its level
EQ-Bench 3 (emotional intelligenc… 1141.5 ▁·▂▁▄●▅▃█▄▁▁ beats 18/62 avg 1034 1.1% typical
AraGen-v3 3C3H score (Inception, … 35.2% ▁▁·▁▂▂▄█●▅▄▄ beats 34/66 hard · avg 29% 1.0% at its level
CPTUB average (SpeakLeash, 0-5 ju… 74.2% ▁▂▂▂▃█▃▆●▄▂▃ beats 67/95 medium · avg 65% 1.0% typical
LiveBench 2024-11-25 (maker) 74.9% ▄··▂·▂·▂●▆▂▂ beats 6/13 medium · avg 61% 1.0% at its level · several runs
Multi-IF (aggregation unspecified… 73.0% ▁▁▁▁▁▁▃●█▄▁▂ beats 16/41 medium · avg 63% 0.9% at its level · several runs
KINA (Knowledge Index of Noah's A… 27.4% ▁▁▂▁▃▄●▆▄▅▂▃ beats 14/39 very hard · avg 21% 0.9% typical
MMMLU 79.0% ▂▁▁▂▃▂▄▄▅●█▃ beats 28/46 medium · avg 64% 0.8% at its level
Open Persian LLM Leaderboard v2 a… 62.8% ▂▄▂▃█▄▃●▃▅▂▂ beats 58/81 medium · avg 57% 0.7% blunt · at its level
Creative Writing v3 (maker) 78.3% ▂····▂●··▆█▆ beats 2/14 easy · avg 69% 0.7% typical
OJBench (maker) 24.1% ▁···▁▃·▅▄●██ beats 11/26 very hard · avg 13% 0.7% typical
KCL-MCQA vanilla 35.0% ▂···█▂●▆██▆▂ beats 6/23 very hard · avg 29% 0.5% typical
LongSocialBench (full-context, 4-… 41.9% ▆▂·▄●·▂▂·▄▄▂ beats 8/16 very hard · avg 36% 0.5% blunt
KOR-Bench 65.4% ●·······█··▄ beats 0/3 medium · avg 55% 0.5% at its level
SimpleQA 8.6% ▄▄▄▆●▆███·▂█ beats 9/32 very hard · avg 6.1% 0.4% typical
RULER 128k 77.5% ▃··▅▄█▅●▇·▃▃ beats 23/33 easy · avg 72% 0.3% typical
IFBench (aggregation unspecified) 35.4% ██●···██···█ beats 2/5 hard · avg 34% 0.3% at its level
EuroEval HellaSwag (English), tes… 84.0% ▃·▁▁▂▁▂▅█●▄▁ beats 81/100 medium · avg 67% 0.3% sharp · at its level · noisy
Chess Puzzles (Epoch) 5.0% ●▁▁▁▁▂▃▃▃▂▁▁ beats 34/137 very hard · avg 5.8% 0.2% sharp · too hard for it · noisy
BeyondAIME 29.0% ▄·●·▂▄··▄▄▄█ beats 2/15 medium · avg 37% 0.2% at its level · noisy
EuroEval Life in the UK (English … 94.1% ▁▁▁▁▁▁▁▂▄█▄● beats 134/146 easy · avg 82% 0.1% noisy
Vectara hallucination leaderboard… 97.2% ▁▁▂▃▄▄█●▃▂▂▁ beats 108/140 very easy · avg 94% 0.1% blunt · maxed out · noisy
Vectara hallucination leaderboard… 94.1% ▁▂▁▂▅█▂▃●▃▁▁ beats 78/98 very easy · avg 90% 0.0% blunt
source results share of the score
Qwen/Qwen3-Next-80B-A3B-Thinking model … 22 20% ▇▇
Qwen/Qwen3-Next-80B-A3B-Instruct model … 24 20% ▇▇
Falcon-H1R: Pushing the Reasoning Front… 12 16% ▇▇
ByteDance-Seed/Seed-OSS-36B-Instruct mo… 18 15% ▇▇
Epoch AI 3 4% ▇
Aider polyglot benchmark runs 1 4% ▇
MUA-RL: Multi-turn User-interacting Age… 1 3% ▇
Qwen (submitted / maker-reported) 1 2% ▇
Arena-Hard-Auto v2.0-Preview leaderboard 2 2% ▇
LMArena leaderboard dataset 1 2% ▇
Korean Canonical Legal Benchmark: Towar… 2 2% ▇
Berkeley Function Calling Leaderboard r… 2 1% ▇
BRIDGE Medical Leaderboard 1 1% ▇
EQ-Bench 3 canonical leaderboard results 1 1% ▇
WebApp1K models leaderboard 1 1% ▇
Inception Arabic/Hindi leaderboards 1 1% ▇
SpeakLeash CPTUB 1 1% ▇
Knowledge Index of Noah's Ark 1 1% ▇
Open Persian LLM Leaderboard v2 1 1% ▇
allenai/Olmo-3.1-32B-Think model card 2 1% ▇
LongSocialBench: Do Long-Context LLMs U… 1 0% ▇
EuroEval leaderboards 2 0% ▇
Vectara hallucination leaderboard 2 0% ▇
prior: pull toward the middle 0% ▇
weighted average = 57.4
Licences: Qwen/Qwen3-Next-80B-A3B-Thinking model card ([apache-2.0](../../licenses/Qwen--Qwen3-Next-80B-A3B-Thinking.txt)) · Qwen/Qwen3-Next-80B-A3B-Instruct model card ([apache-2.0](../../licenses/Qwen--Qwen3-Next-80B-A3B-Instruct.txt)) · Falcon-H1R: Pushing the Reasoning Frontiers with a Hybrid Model for Efficient Test-Time Scaling ([CC BY 4.0](../../licenses/report-2601.02346v1.txt)) · ByteDance-Seed/Seed-OSS-36B-Instruct model card ([apache-2.0](../../licenses/ByteDance-Seed--Seed-OSS-36B-Instruct.txt)) · Epoch AI ([CC BY 4.0](https://creativecommons.org/licenses/by/4.0/)) · Aider polyglot benchmark runs ([Apache-2.0](../../licenses/aider-polyglot-runs-Apache-2.0.txt)) · MUA-RL: Multi-turn User-interacting Agent Reinforcement Learning for agentic tool use ([CC-BY-4.0](https://creativecommons.org/licenses/by/4.0/)) · Arena-Hard-Auto v2.0-Preview leaderboard ([Apache-2.0](../../licenses/arena-hard-auto-Apache-2.0.txt)) · LMArena leaderboard dataset ([CC BY 4.0](https://creativecommons.org/licenses/by/4.0/)) · Korean Canonical Legal Benchmark: Toward Knowledge-Independent Evaluation of LLMs’ Legal Reasoning Capabilities ([CC BY 4.0](../../licenses/cards-final-NOTICE.txt)) · Berkeley Function Calling Leaderboard result archive ([Apache-2.0](../../licenses/bfcl-result-archive-v4-Apache-2.0.txt)) · BRIDGE Medical Leaderboard ([MIT](../../licenses/bridge-medical-leaderboard-MIT.txt)) · EQ-Bench 3 canonical leaderboard results ([MIT](../../licenses/eqbench3-MIT.txt)) · WebApp1K models leaderboard ([MIT](../../licenses/webapp1k-leaderboard-MIT.txt)) · Inception Arabic/Hindi leaderboards ([Apache-2.0](../../licenses/hf-leaderboard-spaces-Apache-2.0.txt)) · SpeakLeash CPTUB ([MIT](../../licenses/hf-leaderboard-spaces-MIT.txt)) · Knowledge Index of Noah's Ark ([CC BY 4.0](https://creativecommons.org/licenses/by/4.0/)) · Open Persian LLM Leaderboard v2 ([Apache-2.0](../../licenses/hf-leaderboard-spaces-Apache-2.0.txt)) · allenai/Olmo-3.1-32B-Think model card ([Apache-2.0](../../licenses/allenai--Olmo-3.1-32B-Think.txt)) · LongSocialBench: Do Long-Context LLMs Understand Online Discussion Threads? ([CC BY 4.0](https://creativecommons.org/licenses/by/4.0/)) · EuroEval leaderboards ([MIT](https://github.com/EuroEval/leaderboards/blob/1a77f1d2f8ba1ba05ab73287fb090d76bf41b2e0/LICENSE)) · Vectara hallucination leaderboard ([Apache-2.0](https://www.apache.org/licenses/LICENSE-2.0))
## How to run it
### Through an API
api (Deep Infra) · $0.13/M blended · $0.08 in / $0.28 out per 1M tokens · [deepinfra.com](https://deepinfra.com/models)
[price source](https://github.com/anomalyco/models.dev/blob/a668908cfe7073faa0363699486fc41a3ff52412/providers/deepinfra/models/Qwen/Qwen3-32B.toml): Standard base token rate from models.dev. Provider-specific tiers, long-context surcharges, caching and promotions may differ. · [context source](https://github.com/anomalyco/models.dev/blob/a668908cfe7073faa0363699486fc41a3ff52412/models/alibaba/qwen3-32b.toml)
### 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 32B (Q4_K_M, needs ~19.8 GB, from [huggingface.co/Qwen](https://huggingface.co/Qwen/Qwen3-32B-GGUF/tree/938a7432affaec9157f883a87164e2646ae17555))
```sh
llama cli -hf Qwen/Qwen3-32B-GGUF:Q4_K_M
```
3. Or start a local server and open the address it prints (http://localhost:9931)
```sh
llama serve -hf Qwen/Qwen3-32B-GGUF:Q4_K_M
```
Other versions: Q8_0 (~34.8 GB) `llama cli -hf Qwen/Qwen3-32B-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 32B (Q4_K_M, needs ~19.8 GB, from [huggingface.co/Qwen](https://huggingface.co/Qwen/Qwen3-32B-GGUF/tree/938a7432affaec9157f883a87164e2646ae17555))
```sh
lms get https://huggingface.co/Qwen/Qwen3-32B-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 (~34.8 GB) `lms get https://huggingface.co/Qwen/Qwen3-32B-GGUF@Q8_0 -y` · 4bit (~17.4 GB, Mac only) `lms get https://huggingface.co/Qwen/Qwen3-32B-MLX-4bit -y` · 8bit (~33.8 GB, Mac only) `lms get https://huggingface.co/Qwen/Qwen3-32B-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 32B (Q4_K_M, needs ~20.18 GB, from [ollama.com](https://ollama.com/library/qwen3/tags))
```sh
ollama run qwen3:32b
```
Other versions: Q8_0 (~39.86 GB) `ollama run qwen3:32b-q8_0`
### 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 32B (4bit, needs ~17.4 GB, Mac only, from [huggingface.co/Qwen](https://huggingface.co/Qwen/Qwen3-32B-MLX-4bit/tree/ceb4c4aad0164033af0c8dd37a0641693d0d6321))
```sh
mlx_lm.chat --model Qwen/Qwen3-32B-MLX-4bit
```
3. Or start an OpenAI-compatible server at http://localhost:8080/v1
```sh
mlx_lm.server --model Qwen/Qwen3-32B-MLX-4bit
```
Other versions: 8bit (~33.8 GB, Mac only) `mlx_lm.chat --model Qwen/Qwen3-32B-MLX-8bit`
Model repository and documentation: [huggingface.co/Qwen](https://huggingface.co/Qwen/Qwen3-32B)
## Quantization evidence
How much of the full-precision result a quantized version kept, measured within one study. Rows from the same study share a table and aren't independent. They don't describe every download of that precision.
precision test full → quantized kept
AutoRound 2-bit (group size 3… WMT24++ X->English COMET (wmt22-c… 0.801 → 0.758 94.7%
AutoRound 2-bit (group size 3… WMT24++ English->X COMET (wmt22-c… 0.751 → 0.592 78.7%
AutoRound 4-bit (group size 1… WMT24++ X->English COMET (wmt22-c… 0.801 → 0.8 99.8%
AutoRound 4-bit (group size 1… WMT24++ English->X COMET (wmt22-c… 0.751 → 0.746 99.3%
AWQ 4-bit (AutoAWQ, zero poin… WMT24++ X->English COMET (wmt22-c… 0.801 → 0.801 99.9%
AWQ 4-bit (AutoAWQ, zero poin… WMT24++ English->X COMET (wmt22-c… 0.751 → 0.747 99.4%
AWQ INT4 (w4, group_size 128,… GPQA (subset not stated in card),… 0.684 → 0.69 100.9%
AWQ INT4 (w4, group_size 128,… LiveBench 2024-11-25, non-thinkin… 0.598 → 0.598 100.0%
AWQ INT4 (w4, group_size 128,… MMLU-Redux, thinking mode (Qwen-r… 0.909 → 0.908 99.9%
AWQ INT4 (w4, group_size 128,… MMLU-Redux, non-thinking mode (Qw… 0.857 → 0.856 99.9%
AWQ INT4 (w4, group_size 128,… LiveBench 2024-11-25, thinking mo… 0.749 → 0.731 97.6%
AWQ INT4 (w4, group_size 128,… AIME 2024, thinking mode (Qwen-re… 0.814 → 0.794 97.5%
AWQ INT4 (w4, group_size 128,… GPQA (subset not stated in card),… 0.546 → 0.531 97.3%
bitsandbytes NF4 (Transformer… WMT24++ X->English COMET (wmt22-c… 0.801 → 0.799 99.8%
bitsandbytes NF4 (Transformer… WMT24++ English->X COMET (wmt22-c… 0.751 → 0.743 98.8%
FP8 block (weights FP8 128x12… GSM8K — published row 'GSM8K (Str… 0.742 → 0.745 100.5%
FP8 block (weights FP8 128x12… MMLU — published row 'MMLU (Acc, … 0.82 → 0.82 100.0%
FP8 block (weights FP8 128x12… OpenLLM v1 average of 6 tasks (AR… 0.747 → 0.745 99.7%
FP8 block (weights FP8 128x12… OpenLLM v2 average of 6 tasks (IF… 0.305 → 0.296 97.1%
FP8 dynamic (weights FP8 per-… GPQA Diamond — published row 'GPQ… 0.667 → 0.687 103.0%
FP8 dynamic (weights FP8 per-… LiveCodeBench — published row 'Li… 0.625 → 0.633 101.4%
FP8 dynamic (weights FP8 per-… OpenLLM v2 average of 6 tasks (MM… 0.421 → 0.427 101.4%
FP8 dynamic (weights FP8 per-… GSM8K — published row 'GSM-8K (5-… 0.876 → 0.883 100.8%
FP8 dynamic (weights FP8 per-… OpenLLM v1 average of 6 tasks (MM… 0.729 → 0.729 100.1%
FP8 dynamic (weights FP8 per-… AIME 2024 — published row 'AIME 2… 0.794 → 0.794 100.0%
FP8 dynamic (weights FP8 per-… MMLU — published row 'MMLU (5-sho… 0.81 → 0.809 99.9%
FP8 dynamic (weights FP8 per-… AIME 2025 — published row 'AIME 2… 0.718 → 0.704 98.1%
GGUF Q2_K with imatrix (llama… WMT24++ X->English COMET (wmt22-c… 0.801 → 0.766 95.5%
GGUF Q2_K with imatrix (llama… WMT24++ English->X COMET (wmt22-c… 0.751 → 0.699 93.0%
GGUF Q4_K_M with imatrix (lla… WMT24++ X->English COMET (wmt22-c… 0.801 → 0.8 99.8%
GGUF Q4_K_M with imatrix (lla… WMT24++ English->X COMET (wmt22-c… 0.751 → 0.75 99.8%
GPTQ-style W2A16 (weight-only… GSM8K accuracy, thinking mode, 32… 0.958 → 0.933 97.4%
GPTQ-style W2A16 (weight-only… MATH-500 accuracy, thinking mode,… 0.93 → 0.65 69.9%
GPTQ-style W2A16 (weight-only… GPQA-Diamond accuracy, thinking m… 0.662 → 0.429 64.9%
GPTQ-style W2A16 (weight-only… AIME 2026 accuracy, thinking mode… 0.7 → 0.167 23.8%
INT4 W4A16 (weight-only INT4,… GPQA Diamond — published row 'GPQ… 0.667 → 0.667 100.0%
INT4 W4A16 (weight-only INT4,… OpenLLM v1 average of 6 tasks (MM… 0.729 → 0.725 99.5%
INT4 W4A16 (weight-only INT4,… MMLU — published row 'MMLU (5-sho… 0.81 → 0.804 99.3%
INT4 W4A16 (weight-only INT4,… GSM8K — published row 'GSM-8K (5-… 0.876 → 0.86 98.1%
INT4 W4A16 (weight-only INT4,… OpenLLM v2 average of 6 tasks (MM… 0.421 → 0.412 97.9%
INT4 W4A16 (weight-only INT4,… AIME 2024 — published row 'AIME 2… 0.794 → 0.773 97.4%
INT4 W4A16 (weight-only INT4,… LiveCodeBench — published row 'Li… 0.625 → 0.596 95.5%
INT4 W4A16 (weight-only INT4,… AIME 2025 — published row 'AIME 2… 0.718 → 0.643 89.5%
NVFP4 W4A4 (FP4 weights and a… GPQA Diamond — published row 'GPQ… 0.629 → 0.645 102.4%
NVFP4 W4A4 (FP4 weights and a… GSM8K — published row 'gsm8k' (Op… 0.742 → 0.747 100.7%
NVFP4 W4A4 (FP4 weights and a… OpenLLM v1 average of 6 tasks (ar… 0.743 → 0.742 99.8%
NVFP4 W4A4 (FP4 weights and a… MMLU — published row 'mmlu' (Open… 0.818 → 0.812 99.3%
NVFP4 W4A4 (FP4 weights and a… OpenLLM v2 average of 6 tasks (BB… 0.557 → 0.544 97.8%
NVFP4 W4A4 (FP4 weights and a… Reasoning average of 3 tasks (AIM… 0.704 → 0.663 94.2%
NVFP4 W4A4 (FP4 weights and a… AIME 2024 — published row 'AIME24… 0.759 → 0.69 90.9%
NVFP4 W4A4 (FP4 weights and a… AIME 2025 — published row 'AIME25… 0.724 → 0.655 90.5%
NVFP4A16 (FP4 weights, group … GSM8K — published row 'GSM8K (8-s… 0.873 → 0.877 100.4%
NVFP4A16 (FP4 weights, group … OpenLLM v1 average of 6 tasks (MM… 0.727 → 0.727 100.0%
NVFP4A16 (FP4 weights, group … MMLU — published row 'MMLU' (Open… 0.809 → 0.806 99.5%
NVFP4A16 (FP4 weights, group … OpenLLM v2 average of 6 tasks (MM… 0.554 → 0.549 99.1%
Each kept value links to its source.
[interactive version](https://whatmodel.sh/model/qwen3-32b) · [JSON](https://whatmodel.sh/api?format=json&all=1)