Skip to content
Leaderboard / Alibaba (Qwen)

Qwen3.5 397B A17B

Alibaba (Qwen) OPEN Apache 2.0
Verdict

Qwen3.5 397B A17B is a solid mid-pack model: #106 of 406 by SIscope score. Strongest at chat quality, weaker at coding. Mid-range on price. Weights are open, so you can run it yourself.

SIscope score
65
#106 of 406
Human-vote rank
#83
of 419 · score 1438
Price per 1M tokens, blended
$1.27
$0.5 in · $3.6 out
Context window
262K
262,144 tokens

Shape

Each spoke is a category, 0 at the centre, 100 at the edge.

CodingAgentsno data yetReasoningMathChatInstructions

Category scores

Percentiles: 90 means better than about 90% of tested models.

Coding57
Agents & toolsNo data yet–
Reasoning & knowledge76
Math63
Chat quality80
Following instructions80

What goes into each category · Compare with Gemini 4 Carbon

Human-vote score over time

Text arena, recent updates.

143014381445Oct 20: 1437.5 (rank #81)Oct 22: 1437.8 (rank #81)Oct 28: 1437.6 (rank #83)Oct 20Oct 28

Coding votes over time

WebDev arena, recent updates.

139514001405Oct 16: 1400.5 (rank #93)Oct 18: 1400.4 (rank #94)Oct 19: 1400.4 (rank #97)Oct 22: 1400.4 (rank #97)Oct 23: 1400.3 (rank #98)Oct 27: 1400.4 (rank #99)Oct 28: 1400.2 (rank #100)Oct 16Oct 28

Independent test runs

Run by Epoch AI on the same harness for every model. Bars show the share of questions answered correctly.

GPQA Diamond86%
PhD-level science multiple choice questions
Better than 71% of 333 tested models. Run Aug 7, 2026.
OTIS Mock AIME 2024-2589%
Competition maths problems in the style of AIME
Better than 71% of 313 tested models. Run Aug 7, 2026.
FrontierMath Tiers 1-329%
Very hard, unpublished research-style maths problems
Better than 36% of 122 tested models. Run Aug 28, 2026.

Human-vote arenas

People compare two anonymous answers and pick the better one. Scores work like chess ratings.

ArenaRankScoreRange (95%)VotesUpdated
Text arena
Everyday chat, blind human votes
#83 of 41914381435–144197,675Oct 28, 2026
WebDev arena
Building web apps, blind human votes
#100 of 15214001395–140521,096Oct 28, 2026
Vision arena
Questions about images, blind human votes
#54 of 16712631257–126832,657Oct 19, 2026
Agent arena
Real multi-step agent sessions
Not on this board yet

Source: Arena (formerly LMArena) leaderboard dataset, CC BY 4.0. Fetched Nov 4, 2026.

Price

Input
$0.5
per 1M tokens
Output
$3.6
per 1M tokens
Blended
$1.27
3 in : 1 out, per 1M
Context window
262K
max tokens per request
Speed
Coming soon
our own tests

List price from the OpenRouter models API (qwen/qwen3.5-397b-a17b), Nov 4, 2026. Providers can charge differently.