whatmodel.sh / Qwen3-Max (2026-01-23)

Qwen3-Max (2026-01-23)

+ contribute
Update model informationAdd a place to run itAdd a test resultPropose a benchmarkCorrect a benchmark scoreEdit data on GitHub

Qwen3-Max (2026-01-23) is a closed-weights language model from Alibaba, released January 2026. Licence: proprietary.

Catalogue identity notes

Comparison-only, proposed so the tau2-bench-leaderboard-qwen telecom run counts. Dated Alibaba Cloud Model Studio snapshot 'qwen3-max-2026-01-23' (release log 2026-01-23: 'Compared with the snapshot as of September 23, 2025, the Qwen-3 series Max model in this release achieves an effective integration of thinking and non-thinking modes'); marketed as Qwen3-Max-Thinking (qwen.ai blog id=qwen3-max-thinking). Distinct from qwen3-max-2025-09-23 and qwen3-max-preview.

69.8±0.9comparison only: no current run option verifiedfrom 40 benchmark results across 2 sources

$ how it compares

The LLMs ranked either side of Qwen3-Max (2026-01-23), out of 828 with a score, then more from Alibaba.

#modelscorewhereRAMreleased
92Kimi K2.5 Moonshot AI70±1api $1.20/M +local~160GB2026-01
93GLM-5 Z.ai70±1api $1.55/M +local~119GB2026-02
94Qwen3-Max (2026-01-23) Alibaba70±1comparison only2026-01
95MiMo-V2.5-Pro (1T-A42B) Xiaomi70±1api $0.65/M +local~163GB2026-04
96Inkling (975B-A41B) Thinking Machines70±1api $1.73/M +local~156GB2026-07
more from Alibaba
18Qwen3.8 Max (2.4T-A95B) Alibaba78±1api $3.00/M +local~380GB2026-08
34Qwen3.8 Max (0902) Alibaba76±1api $3.00/M2026-09
37Qwen3.8-Flash-Next Alibaba75±1local~59GB2026-08

see all LLMs · find the ones that fit your machine

$ what it's good at

agent tasks▁▁▂▆█▆▅▄▄▃▂▁ #91 of 340 · 38% of its score · 14 tests

Beats Grok 4.20 (0309 reasoning), GPT-5.1, Grok 4.3 here, though they rank higher overall. Trails Hy3 preview (295B-A21B), MiniMax-M2.7, Step 3.5 Flash (196B-A11B) here, though they rank lower overall.

testresultvs the others in its rundifficultycountswhy
τ²-bench aggregate lab84.6% ✎▄····█▄▄···▄ beats 1/5medium · avg 40%▇▇ 5.6%typical
SWE-bench Verified lab75.3% ✎▁▁▁▁▁▁▂▄█▆▁▁ beats 102/131hard · avg 23%▇▇ 4.4%sharp · at its level · small run
τ²-bench Telecom lab98.3% ✎▁·▅█▆▅▂▃▃▂▃▁ beats 85/89hard · avg 34%▇▇ 4.1%sharp · small run
BrowseComp-ZH lab60.9% ✎▁▁▁▁▁▃·▁▄▅▄█ beats 15/23hard · avg 19%▇ 3.3%at its level · small run
SWE-bench Multilingual lab66.7% ✎▁··▁▃▂▅▇█▃▃▁ beats 27/60hard · avg 26%▇ 2.9%at its level · small run
τ²-bench Retail lab79.4% ✎▁····▂▁▃▆█▃▁ beats 41/44medium · avg 39%▇ 2.5%small run
Terminal-Bench 2.0 lab22.5% ✎▂▂▅▅████▆▂▃▃ beats 2/36very hard · avg 8.5%▇ 2.3%at its level · small run
BrowseComp lab53.9% ✎▂·▄▁▂▃▆▄█▂▃▄ beats 21/42hard · avg 20%▇ 2.3%at its level · small run
6 more tests
testresultvs the others in its rundifficultycountswhy
τ²-bench Airline lab69.0% ✎▁▁·▂▁▃▂▁█▅▂▃ beats 40/42hard · avg 30%▇ 2.0%at its level · small run
Humanity's Last Exam with tools lab49.8% ✎▂▂▂▁▄▂▃▆█▇▂▆ beats 30/54hard · avg 17%▇ 2.0%at its level · small run
MCPMark lab33.5% ✎▃·····█··▃█▃ beats 5/8very hard · avg 9.9%▇ 2.0%at its level · small run
BFCL V4 (December 2025 archive) lab67.7% ✎▁▁▁▂▇▆▆▆█▂▄▂ beats 101/111medium · avg 39%▇ 1.6%at its level · small run
Toolathlon (Tool Decathlon) lab18.8% ✎▃▂▄▂▄▄▆▂█▆██ beats 0/27very hard · avg 10%▇ 1.5%at its level · small run
WideSearch lab57.9% ✎█·▂▆··▂···▂▂ beats 0/10hard · avg 35%▇ 1.4%at its level · small run
coding▁▁▁▃▅▇█▇▅▄▃▁ #67 of 524 · 10% of its score · 3 tests

Beats DeepSeek V4 Flash Vision (Exp), GPT-5.2, GPT-5 here, though they rank higher overall. Trails MiMo-V2.5-Pro (1T-A42B) here, though they rank lower overall.

testresultvs the others in its rundifficultycountswhy
SWE-bench Verified lab75.3% ✎▁▁▁▁▁▁▂▄█▆▁▁ beats 102/131hard · avg 23%▇▇ 4.4%sharp · at its level · small run
SWE-bench Multilingual lab66.7% ✎▁··▁▃▂▅▇█▃▃▁ beats 27/60hard · avg 26%▇ 2.9%at its level · small run
LiveCodeBench v6 (Feb–May 2025) lab85.9% ✎▁▁▃▁█▅▄▇▅▅▃█ beats 36/38medium · avg 46%▇ 2.5%small run
chat & writing▁▁▁▃▅▄▆█▇▇▇▃ #106 of 463 * · 1.5% of its score · 1 test

Beats LongCat 2.0 (1.6T-A48B), Inkling Small (266B-A12B) here, though they rank higher overall. Trails Gemma 4 12B, Kimi K2 Instruct (0905), Gemma 4 26B-A4B here, though they rank lower overall.

testresultvs the others in its rundifficultycountswhy
IFEval (aggregation unspecified) lab93.4% ✎▁▁▃▁▂▄▄▆█▃▃▁ beats 132/141easy · avg 79%▇ 1.5%maxed out · small run
math▁▂▃▅▇▅▅█▄▁▁▁ #51 of 462 · 15% of its score · 5 tests

Beats LongCat 2.0 (1.6T-A48B), Qwen3.6 Max Preview, Claude Sonnet 4.6 here, though they rank higher overall. Trails A.X K2 (688B-A33B) here, though they rank lower overall.

testresultvs the others in its rundifficultycountswhy
AIME 2026 lab93.3% ✎▁▁▁▂▁▂▆█▆▂·▁ beats 35/49medium · avg 52%▇▇ 4.6%sharp · maxed out · small run
IMO-AnswerBench lab83.9% ✎▁▁·▁▁▁▄▅█▄▂▃ beats 31/42medium · avg 36%▇ 3.7%small run
HMMT February 2025 lab98.0% ✎▁▁▁▃▁▅▅▃██▁▄ beats 30/31easy · avg 70%▇ 2.9%maxed out · small run
HMMT November 2025 lab94.7% ✎▁▁····▁▅▃█▇▁ beats 22/23easy · avg 71%▇ 2.2%small run
PolyMATH lab64.7% ✎▁···▃▁·▇█▄▄▄ beats 20/23hard · avg 32%▇ 1.9%at its level · small run
web search▂·▂▇▅█▃▄▁▂▁▁ #75 of 92 · 10% of its score · 5 tests

Beats DeepSeek V4 Flash (Preview, Apr 2026) here, though they rank higher overall. Trails GPT-4.1 nano, GPT-4o mini (2024-07-18), Mistral Small 3.1 24B (2503) here, though they rank lower overall.

testresultvs the others in its rundifficultycountswhy
BrowseComp-ZH lab60.9% ✎▁▁▁▁▁▃·▁▄▅▄█ beats 15/23hard · avg 19%▇ 3.3%at its level · small run
BrowseComp lab53.9% ✎▂·▄▁▂▃▆▄█▂▃▄ beats 21/42hard · avg 20%▇ 2.3%at its level · small run
Humanity's Last Exam with tools lab49.8% ✎▂▂▂▁▄▂▃▆█▇▂▆ beats 30/54hard · avg 17%▇ 2.0%at its level · small run
SealQA Seal-0 lab46.9% ✎▃····▃·····█ beats 2/4hard · avg 19%▇ 1.5%at its level · small run
WideSearch lab57.9% ✎█·▂▆··▂···▂▂ beats 0/10hard · avg 35%▇ 1.4%at its level · small run
other tests19 not in a task · 41% of its score
testresultvs the others in its rundifficultycountswhy
GPQA (subset unspecified) lab87.4% ✎▂▁▂▂▄▃▆▃▅█▄▃ beats 100/104hard · avg 47%▇▇ 4.9%sharp · small run
LongBench v2 lab60.6% ✎▂·▁·▂▂▁▃▄▁█▄ beats 21/28very hard · avg 34%▇ 3.3%at its level · small run
Humanity's Last Exam lab30.2% ✎▁▁▁▃▆▆▇██▆▁▃ beats 71/104very hard · avg 5.5%▇ 3.1%at its level · small run
C-Eval lab93.7% ✎▁···▁▁▂▁▂▂▃█ beats 33/33medium · avg 72%▇ 3.0%small run
MMLU-Pro lab85.7% ✎▁▁·▁▁▁▃▂▄▅█▁ beats 175/188medium · avg 55%▇ 2.9%small run
INCLUDE lab82.3% ✎▂····▂▂▄▄▄█▄ beats 33/36medium · avg 55%▇ 2.7%small run
MMLU-Redux lab92.8% ✎▁▁▃▁▂▃▄▅▇▅█▄ beats 49/63easy · avg 77%▇ 2.5%maxed out · small run
VitaBench lab40.9% ✎▄··▂▄▆▆█▄▄▄▂ beats 19/21very hard · avg 12%▇ 2.4%at its level · small run
11 more tests
testresultvs the others in its rundifficultycountswhy
MMMLU lab84.4% ✎▂▁▁▂▃▂▄▄▅██▃ beats 35/46medium · avg 62%▇ 2.3%small run
IFBench (aggregation unspecified) lab70.9% ✎▃▇▂▄▅▅▂▄▄▆█▇ beats 57/80medium · avg 40%▇ 1.9%at its level · small run
Global PIQA lab86.0% ✎▃▃··▃··▃▃▅█▅ beats 7/11medium · avg 71%▇ 1.9%small run
MultiChallenge lab63.3% ✎▁·▁·▁▃▆▅▃▁█▅ beats 25/29hard · avg 30%▇ 1.9%at its level · small run
SuperGPQA lab67.3% ✎▂▁▂▂▅▆▃▇▇▇█▃ beats 45/46medium · avg 37%▇ 1.8%at its level · small run
MAXIFE lab84.0% ✎▂·▂·▂···▆·▄█ beats 7/11medium · avg 61%▇ 1.8%small run
MMLU-ProX lab78.5% ✎▁·▁▁▁▂▁▃▂█▅▄ beats 18/35medium · avg 56%▇ 1.7%small run
DeepPlanning lab28.7% ✎▂·····▂··▄█▄ beats 8/9very hard · avg 9.7%▇ 1.5%at its level · small run
NOVA-63 lab54.2% ✎▃▃·▃···▃█▃▃█ beats 5/11hard · avg 35%▇ 1.3%blunt · at its level · small run
WMT24++ lab77.6% ✎▁·▁▁·▁·▁▂▆█▃ beats 12/27easy · avg 69%▇ 0.3%blunt · small run
SecCodeBench lab57.5% ✎█··········█ beats 0/1—not counted yet: no other model in the same run to compare it with

$ where to run it

No verified API or download recorded yet · add one

$ how the score was worked out · Qwen/Qwen3.5-397B-A17B model… 91%, τ²-bench leaderboard: Qwen (A… 9%

Each result implies a score on its own, after adjusting for how hard its test is and who ran it. The 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's added up by who ran the tests. How it works

sourceresultsshare of the scorelicence
Qwen/Qwen3.5-397B-A17B model card · lab-reported37▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇ 91%apache-2.0
τ²-bench leaderboard: Qwen (Alibaba) own submissions · lab-reported3▇▇▇ 9%MIT
prior · pull toward the middle▇ 0%
weighted average= 69.8

not counted (1)

Qwen/Qwen3.5-397B-A17B model cardSecCodeBench [correct]57.5%fewer than three models have this test yet, or this source ran it for this model alone

$ improve this page

Update model informationAdd a place to run itAdd a test resultPropose a benchmarkCorrect a benchmark scoreEdit data on GitHub

Use a short GitHub form or edit the data through a pull request. Every number needs a source; here's which ones we can use.

data updated · sources and licences · json