Qwen3 235B 2507 vs Qwen3 235B-A22B

Change models
IndexAll categories, out of 100.
42.3 ±8.1 best
38.1 ±8.8
GapIndex points below the top.
Top
−4.2 · no clear gap
CoverageIndex weight that has results.
70% best
55%
SpeedOutput tokens per second.
48/s best
44/s
EloLMArena rating and rank.
1419 best
1366
ReleasedMonth of release.
2025-07
2025-04
Input / 1MUS dollars per 1M input tokens.
$0.14 best
$0.455
Output / 1MUS dollars per 1M output tokens.
$0.80 best
$1.82
1K chats2K tokens in, 500 out, each.
$0.68 best
$1.82
1K documents20K tokens in, 1K out, each.
$3.60 best
$10.92
1K agent tasks200K tokens in, 20K out, each.
$44.00 best
$127.40
ContextMaximum tokens in one request.
262K best
131K
AgenticMulti-step tasks with tools.
—
—
CodingCode writing and repair.
26.9
44.0 best
ReasoningLogic problems and puzzles.
35.9
—
MultimodalTasks with images and text.
—
—
KnowledgeFacts and expert knowledge.
51.2 best
41.1
MultilingualTasks in many languages.
48.8 best
42.0
InstructionTasks with strict prompt rules.
41.8 best
30.7
MathMath problems.
37.2 best
32.2

Benchmarks

3 shared
Only Qwen3 235B 2507: 9Only Qwen3 235B-A22B: 21
Qwen3 235B 2507
Qwen3 235B-A22B
77.5% best74 / 141
62.9%108 / 141
88.7% best29 / 70
83.2%46 / 70
70.3% best36 / 53
24.7%50 / 53