DeepSeek V3.2 Thinking vs GLM-4.7

Change models
IndexAll categories, out of 100.
49.0 ±8.9
49.5 ±4.6 best
GapIndex points below the top.
−0.5 · no clear gap
Top
CoverageIndex weight that has results.
65%
85% best
SpeedOutput tokens per second.
—
72/s
EloLMArena rating and rank.
—
1435
ReleasedMonth of release.
—
2025-12
Input / 1MUS dollars per 1M input tokens.
—
$0.40
Output / 1MUS dollars per 1M output tokens.
—
$1.75
1K chats2K tokens in, 500 out, each.
—
$1.68
1K documents20K tokens in, 1K out, each.
—
$9.75
1K agent tasks200K tokens in, 20K out, each.
—
$115.00
ContextMaximum tokens in one request.
—
205K
AgenticMulti-step tasks with tools.
47.4
48.3 best
CodingCode writing and repair.
48.8
52.5 best
ReasoningLogic problems and puzzles.
—
48.2
MultimodalTasks with images and text.
—
—
KnowledgeFacts and expert knowledge.
50.8 best
50.7
MultilingualTasks in many languages.
—
—
InstructionTasks with strict prompt rules.
—
51.8
MathMath problems.
51.9 best
48.6

Benchmarks

15 shared
Only DeepSeek V3.2 Thinking: 1Only GLM-4.7: 31
DeepSeek V3.2 Thinking
GLM-4.7
76.1%112 / 149
83.4% best55 / 149
TaxEval v2Finance
68.2%100 / 143
68.8% best99 / 143
80.7%65 / 142
82.2% best53 / 142
80.3% best66 / 135
80.0%68 / 135
MMLU ProKnowledge
84.9% best59 / 135
82.7%76 / 135
CorpFin v2Finance
51.0% best105 / 131
46.4%113 / 131
AIMEMath
84.6%39 / 95
93.3% best15 / 95
MedQAHealthcare
93.9% best20 / 95
93.7%21 / 95
SWE-benchCoding
67.6%69 / 87
69.4% best66 / 87
MGSMMultilingual
86.0%65 / 74
88.2% best53 / 74
36.0%41 / 66
38.2% best36 / 66
IOI v1Coding
10.7% best34 / 61
7.6%38 / 61
55.4% best34 / 53
54.9%36 / 53
2.3%49 / 51
46.0% best36 / 51
40.0%22 / 47
50.0% best10 / 47