DeepSeek V3.2 vs Gemma 4 26B A4B

Change models
IndexAll categories, out of 100.
44.0 ±4.9
44.2 ±3.9 best
GapIndex points below the top.
−0.2 · no clear gap
Top
CoverageIndex weight that has results.
85%
100% best
SpeedOutput tokens per second.
19/s
—
EloLMArena rating and rank.
1425
1434 best
ReleasedMonth of release.
2025-12
2026-04
Input / 1MUS dollars per 1M input tokens.
$0.26
Free best
Output / 1MUS dollars per 1M output tokens.
$0.38
Free best
1K chats2K tokens in, 500 out, each.
$0.71
Free best
1K documents20K tokens in, 1K out, each.
$5.58
Free best
1K agent tasks200K tokens in, 20K out, each.
$59.60
Free best
ContextMaximum tokens in one request.
164K
256K best
AgenticMulti-step tasks with tools.
42.9 best
34.0
CodingCode writing and repair.
48.6 best
45.1
ReasoningLogic problems and puzzles.
42.7 best
41.9
MultimodalTasks with images and text.
—
49.1
KnowledgeFacts and expert knowledge.
43.1
46.5 best
MultilingualTasks in many languages.
—
36.3
InstructionTasks with strict prompt rules.
34.8
56.5 best
MathMath problems.
46.6
56.5 best

Benchmarks

8 shared
Only DeepSeek V3.2: 30Only Gemma 4 26B A4B: 21
DeepSeek V3.2
Gemma 4 26B A4B
11.2%138 / 222
19.3% best114 / 222
45.7%163 / 217
65.7% best134 / 217
77.3% best96 / 217
73.2%108 / 217
0.9% best124 / 214
0.0%150 / 214
68.4%103 / 200
82.2% best81 / 200
75.1%118 / 199
79.2% best102 / 199
78.9% best71 / 149
43.6%102 / 149
49.0%80 / 146
72.4% best36 / 146