[whatmodel.sh](../../README.md) / Qwen3.8 Max (2.4T-A95B)
# Qwen3.8 Max (2.4T-A95B)
Alibaba · released 2026-08 · 2400B params (95B active) · qwen3.8-max-license
**score 76.4** ±0.6 · #23 of 732 · from 24 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 Qwen3.8 Max (2.4T-A95B) among every model on that task or test, worst to best.
task vs the others rank counts tests
agent tasks ▁▁▃▆█▄▅▅▄●▂▁ #29 of 313 14% 3
coding ▁▁▁▃▅▇█▇▅▄●▁ #14 of 469 11% 3
chat & writing ▁▁▁▃▆▄▆█▆▇▇● #19 of 402* 3.6% 1
math ▁▁▃▅█▆▄█▄●▁▁ #21 of 431 51% 9
vision & documents ▁▁▂▂▁▃▄▂▆▇█● #3 of 153* 5.5% 2
web search ▁▁▁▅▄█●▄·▁·▁ #20 of 84* 1.3% 1
agent tasks · #29 of 313 · 14% of its score
Beats [Hy4 preview (770B-A49B)](hy4-preview.md), [GPT-5.6 Terra](gpt-5-6-terra.md) here, though they rank higher overall. Trails [Grok 4.7](grok-4-7.md), [GLM-5.3](glm-5-3.md), [GPT-6 Luna](gpt-6-luna.md) here, though they rank lower overall.
test result vs the others in its run difficulty counts why
Arena Agent (October 2026 aggrega… 0.0045 IPS ▁▁▁▃▃▂●▁▂▂▁▁ beats 25/59 avg -0.379 IPS 9.2% sharp
GameLogicBench solve rate (72 God… 44.4% █▄·█·▄·██●·▄ beats 10/11 very hard · avg 5.1% 2.6% at its level
τ³-bench banking_knowledge (v1.0.… 55.1% ▃▃▂▃·▆▃▂█▅▅● beats 25/25 very hard · avg 4.3% 2.6% at its level
coding · #14 of 469 · 11% of its score
Beats [GPT-5.6 Sol](gpt-5-6-sol.md), [Claude Fable 5](claude-fable-5.md), [Muse Wonder](muse-wonder.md) here, though they rank higher overall. Trails [Qwen3.8 Max (0902)](qwen3-8-max-0902.md) here, though they rank lower overall.
test result vs the others in its run difficulty counts why
Arena WebDev (Code Arena) 1671.9 ▁▁▃▂▅█▃▆▅●▁▂ beats 116/129 avg 981 6.3% sharp
GameLogicBench solve rate (72 God… 44.4% █▄·█·▄·██●·▄ beats 10/11 very hard · avg 5.1% 2.6% at its level
Luau Leetcode fully solved (Torpe… 77.4% ▁▁▁▁·▁▂▃▅▃●▇ beats 27/49 hard · avg 17% 2.2% sharp
chat & writing · #19 of 402 · 3.6% of its score
Beats [GPT-6 Astra](gpt-6-astra.md), [GPT-6.1 Sol](gpt-6-1-sol.md), [Claude Sonnet 5.5](claude-sonnet-5-5.md) here, though they rank higher overall. Trails [Gemini 3.6 Flash](gemini-3-6-flash.md), [Muse Spark 1.2](muse-spark-1-2.md), [Muse Spark 1.1](muse-spark-1-1.md) here, though they rank lower overall.
test result vs the others in its run difficulty counts why
LMArena (text) 1482.4 ▁▁▁▂▃▃▃▆▆▇█● beats 365/385 avg 1260 3.6% blunt
math · #21 of 431 · 51% of its score
Beats [DeepSeek V4.1 Flash 1017](deepseek-v4-1-flash-1017.md) here, though they rank higher overall. Trails [GPT-5.6 Luna](gpt-5-6-luna.md), [GPT-6 Luna](gpt-6-luna.md), [GPT-5.4](gpt-5-4.md) here, though they rank lower overall.
test result vs the others in its run difficulty counts why
FrontierMath tier 4 v2 46.3% ▁▂▁▂█●▂▃▂▁▁▁ beats 47/67 very hard · avg 0.1% 16% sharp · at its level
FrontierMath tiers 1–3 v2 74.7% ▂··▁▁▃▄▄█●▄▃ beats 66/86 very hard · avg 1.2% 13% sharp
MathArena ArXivMath April 2026 (4… 59.2% ▁·····▅█●▄·▁ beats 13/20 very hard · avg 7.6% 7.0% sharp · at its level
MathArena ArXivMath June 2026 (48… 75.7% ▂···▄·██●▆▄▂ beats 13/19 very hard · avg 5.1% 7.0% sharp · at its level
MathArena BrokenArXiv April 2026 … 45.5% ▂▂▆█▃●▂▂▂··▂ beats 13/18 very hard · avg 0.8% 2.9% sharp · at its level · noisy
MathArena ArXivMath May 2026 (40 … 69.2% ▁··▁▄█●▁·▃·▁ beats 13/17 very hard · avg 11% 2.3% at its level
MathArena ArXivMath August 2026, … 57.9% █▄·●·······█ beats 3/5 very hard · avg 5.9% 1.4% sharp · at its level
MathArena BrokenArXiv August 2026… 69.6% ▄▄·····▄·●·█ beats 3/5 very hard · avg 6.9% 0.6% at its level · noisy
OTIS mock AIME 2024–2025 99.4% ▃▃▄▃▃▃▇▇█▃▂● beats 186/197 very hard · avg 7.2% 0.3% sharp · maxed out · noisy
vision & documents · #3 of 153 · 5.5% of its score
Beats [GPT-6 Astra](gpt-6-astra.md), [Claude Opus 5.5](claude-opus-5-5.md), [GPT-6.1 Sol](gpt-6-1-sol.md) here, though they rank higher overall.
test result vs the others in its run difficulty counts why
CharXiv (reasoning questions) (ma… 88.4% ▄▂▄▆▄▄▆··▄█● beats 21/21 medium · avg 38% 3.1% typical
Arena vision (style controlled) 1300.0 ▁▁▂▂▂▂▃▄▅▅█● beats 151/152 avg 1061 2.4% blunt
web search · #20 of 84 · 1.3% of its score
Beats [GPT-6 Sol](gpt-6-sol.md), [Claude Fable 5](claude-fable-5.md), [Claude Mythos Preview (limited preview)](claude-mythos-preview.md) here, though they rank higher overall. Trails [GPT-4o mini (2024-07-18)](gpt-4o-mini-2024-07-18.md), [Mistral Small 3.1 24B (2503)](mistral-small-3-1.md), [GPT-4o (2024-08-06)](gpt-4o-2024-08-06.md) here, though they rank lower overall.
test result vs the others in its run difficulty counts why
Arena search systems (style contr… 1229.5 ▁·▁▁▂▂▂█▆▃●▂ beats 33/36 avg 1069 1.3% blunt
other tests (6)
test result vs the others in its run difficulty counts why
Mystery Game Puzzles (Epoch) 38.0% █··▂▃▇▄●▂▄▂▁ beats 61/77 very hard · avg 9.4% 6.8% sharp · at its level · noisy
Chess Puzzles (Epoch) 29.0% █▁▁▁▁▂▃●▃▂▁▁ beats 102/137 very hard · avg 5.8% 3.7% sharp · at its level · noisy
GPQA Diamond 92.7% ▁▁▁▁▂▄▆▄▅▇█● beats 191/211 hard · avg 39% 1.9% sharp
SimpleQA Verified 45.8% ▁▁▃▂▃▃▆●█▄▂▅ beats 44/81 very hard · avg 8.0% 1.4% at its level
Icelandic LLM Leaderboard average… 44.1% ▁▁▃▂▂▄▁●▄▄▄█ beats 16/38 very hard · avg 14% 1.3% at its level
RobloxQA v2.0 multiple choice (To… 95.4% ▁▁▃▁█▄▃●▆▂▂▂ beats 29/49 easy · avg 77% 0.8% maxed out
source results share of the score
Epoch AI 7 43% ▇▇▇▇▇
LMArena leaderboard dataset 5 23% ▇▇▇
MathArena ArXivMath / BrokenArXiv / ArX… 6 21% ▇▇▇
Qwen (submitted / maker-reported) 1 3% ▇
Roblox LLM Leaderboard 2 3% ▇
GameLogicBench: Evaluating Coding Agent… 1 3% ▇
τ²/τ³-bench leaderboard: Sierra's own e… 1 3% ▇
Miðeind Icelandic LLM Leaderboard 1 1% ▇
prior: pull toward the middle 0% ▇
weighted average = 76.4
Licences: Epoch AI ([CC BY 4.0](https://creativecommons.org/licenses/by/4.0/)) · LMArena leaderboard dataset ([CC BY 4.0](https://creativecommons.org/licenses/by/4.0/)) · MathArena ArXivMath / BrokenArXiv / ArXivLean model-output datasets ([CC BY-SA 4.0](../../licenses/matharena-sharealike-NOTICE.txt)) · Roblox LLM Leaderboard ([MIT](../../licenses/roblox-llm-leaderboard-MIT.txt)) · GameLogicBench: Evaluating Coding Agents on Runtime Game Logic with Tick-Level State Assertions ([CC BY 4.0](https://creativecommons.org/licenses/by/4.0/)) · τ²/τ³-bench leaderboard: Sierra's own evaluation runs ([MIT](https://opensource.org/license/mit)) · Miðeind Icelandic LLM Leaderboard ([CC BY 4.0](../../licenses/mideind-icelandic-llm-leaderboard-CC-BY-4.0.txt))
## How to run it
### Through an API
api (Deep Infra) · $3.00/M blended · $2 in / $6 out per 1M tokens · [deepinfra.com](https://deepinfra.com/models)
[price source](https://github.com/anomalyco/models.dev/blob/a668908cfe7073faa0363699486fc41a3ff52412/providers/deepinfra/models/Qwen/Qwen3.8-2.4T-A95B.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/qwen3.8-2.4t-a95b.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 Qwen3.8 Max (2.4T-A95B) (ud-iq4_xs, needs ~1440.5 GB, from [huggingface.co/unsloth](https://huggingface.co/unsloth/Qwen3.8-2.4T-A95B-GGUF/tree/main))
```sh
llama cli -hf unsloth/Qwen3.8-2.4T-A95B-GGUF:UD-IQ4_XS
```
3. Or start a local server and open the address it prints (http://localhost:9931)
```sh
llama serve -hf unsloth/Qwen3.8-2.4T-A95B-GGUF:UD-IQ4_XS
```
Other versions: ud-iq1_m (~360.5 GB) `llama cli -hf unsloth/Qwen3.8-2.4T-A95B-GGUF:UD-IQ1_M`
### 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 Qwen3.8 Max (2.4T-A95B) (ud-iq4_xs, needs ~1440.5 GB, from [huggingface.co/unsloth](https://huggingface.co/unsloth/Qwen3.8-2.4T-A95B-GGUF/tree/main))
```sh
lms get https://huggingface.co/unsloth/Qwen3.8-2.4T-A95B-GGUF@UD-IQ4_XS -y
```
3. Then run lms ls to see its name, and lms chat followed by that name
Other versions: ud-iq1_m (~360.5 GB) `lms get https://huggingface.co/unsloth/Qwen3.8-2.4T-A95B-GGUF@UD-IQ1_M -y`
Model repository and documentation: [huggingface.co/Qwen](https://huggingface.co/Qwen/Qwen3.8-2.4T-A95B)
[interactive version](https://whatmodel.sh/model/qwen3-8-2-4t-a95b) · [JSON](https://whatmodel.sh/api?format=json&all=1)