GLM-5.1 vs Grok 4.20 (Reasoning)

Change models
IndexAll categories, out of 100.
55.3 ±4.1 best
55.0 ±5.5
GapIndex points below the top.
Top
−0.3 · no clear gap
CoverageIndex weight that has results.
85%
90% best
SpeedOutput tokens per second.
33/s
—
EloLMArena rating and rank.
1463
—
ReleasedMonth of release.
2026-04
—
Input / 1MUS dollars per 1M input tokens.
$1.19
—
Output / 1MUS dollars per 1M output tokens.
$3.74
—
1K chats2K tokens in, 500 out, each.
$4.25
—
1K documents20K tokens in, 1K out, each.
$27.54
—
1K agent tasks200K tokens in, 20K out, each.
$312.80
—
ContextMaximum tokens in one request.
205K
—
AgenticMulti-step tasks with tools.
58.3 best
48.5
CodingCode writing and repair.
56.3 best
51.2
ReasoningLogic problems and puzzles.
50.4
67.7 best
MultimodalTasks with images and text.
—
63.0
KnowledgeFacts and expert knowledge.
51.5
55.5 best
MultilingualTasks in many languages.
—
—
InstructionTasks with strict prompt rules.
59.4
—
MathMath problems.
54.2
57.0 best

Benchmarks

18 shared
Only GLM-5.1: 42Only Grok 4.20 (Reasoning): 10
GLM-5.1
Grok 4.20 (Reasoning)
84.4% best33 / 149
77.7%107 / 149
TaxEval v2Finance
71.2%80 / 143
74.1% best41 / 143
81.4%62 / 142
84.3% best39 / 142
84.5%50 / 135
88.6% best34 / 135
MMLU ProKnowledge
86.9% best37 / 135
86.3%43 / 135
CorpFin v2Finance
64.5% best35 / 131
63.7%40 / 131
31.5% best88 / 130
4.1%117 / 130
MedScribeHealthcare
72.3% best91 / 111
63.4%105 / 111
MedCodeHealthcare
41.6% best53 / 108
32.2%91 / 108
27.9% best68 / 98
13.9%84 / 98
AIMEMath
91.9%21 / 95
96.5% best6 / 95
SWE-benchCoding
76.4% best37 / 87
72.2%56 / 87
44.8% best57 / 79
28.5%73 / 79
25.8% best47 / 77
0.3%76 / 77
56.9% best48 / 75
44.2%64 / 75
53.9% best16 / 66
40.4%32 / 66
51.6%47 / 53
54.4% best37 / 53
57.7% best10 / 51
52.3%24 / 51