Claude Haiku 4.5 vs Qwen3.5-4B

Change models
Qwen3.5-4BAlibaba
IndexAll categories, out of 100.
43.0 ±4.2
43.0 ±7.1
GapIndex points below the top.
Top
Top
CoverageIndex weight that has results.
95% best
80%
SpeedOutput tokens per second.
81/s best
33/s
EloLMArena rating and rank.
1396
—
ReleasedMonth of release.
2025-10
—
Input / 1MUS dollars per 1M input tokens.
$1.00
—
Output / 1MUS dollars per 1M output tokens.
$5.00
—
Batch output / 1MBatch price per 1M output tokens.
$2.50
—
1K chats2K tokens in, 500 out, each.
$4.50
—
1K documents20K tokens in, 1K out, each.
$25.00
—
1K agent tasks200K tokens in, 20K out, each.
$300.00
—
ContextMaximum tokens in one request.
200K
262K best
AgenticMulti-step tasks with tools.
48.5
63.6 best
CodingCode writing and repair.
49.7
—
ReasoningLogic problems and puzzles.
38.0
43.5 best
MultimodalTasks with images and text.
42.7 best
35.5
KnowledgeFacts and expert knowledge.
36.0
41.0 best
MultilingualTasks in many languages.
44.4 best
35.5
InstructionTasks with strict prompt rules.
—
37.6
MathMath problems.
40.3
41.2 best

Benchmarks

9 shared
Only Claude Haiku 4.5: 29Only Qwen3.5-4B: 8
Claude Haiku 4.5
Qwen3.5-4B
EuroEval PolishMultilingual
49.5% best71 / 210
44.2%95 / 210
51.3%121 / 200
55.8% best115 / 200
EuroEval SwedishMultilingual
53.7% best61 / 196
48.9%74 / 196
EuroEval GermanMultilingual
47.8% best48 / 192
39.7%84 / 192
53.4% best58 / 183
48.7%75 / 183
EuroEval FrenchMultilingual
57.6% best53 / 179
51.5%83 / 179
EuroEval SpanishMultilingual
49.7% best45 / 177
36.8%89 / 177
EuroEval DutchMultilingual
53.4% best53 / 176
44.6%86 / 176
EuroEval ItalianMultilingual
53.1%53 / 175
53.3% best52 / 175