Claude Opus 4 vs GPT-OSS 120B

Change models
Claude Opus 4Anthropic
IndexAll categories, out of 100.
43.6 ±5.3 best
43.5 ±4.2
GapIndex points below the top.
Top
−0.1 · no clear gap
CoverageIndex weight that has results.
90%
90%
SpeedOutput tokens per second.
—
224/s
EloLMArena rating and rank.
1366 best
1365
ReleasedMonth of release.
—
2025-08
Input / 1MUS dollars per 1M input tokens.
—
$0.09
Output / 1MUS dollars per 1M output tokens.
—
$0.36
1K chats2K tokens in, 500 out, each.
—
$0.36
1K documents20K tokens in, 1K out, each.
—
$2.16
1K agent tasks200K tokens in, 20K out, each.
—
$25.20
ContextMaximum tokens in one request.
—
131K
AgenticMulti-step tasks with tools.
50.2 best
41.3
CodingCode writing and repair.
46.9 best
42.5
ReasoningLogic problems and puzzles.
38.3
40.0 best
MultimodalTasks with images and text.
50.0
—
KnowledgeFacts and expert knowledge.
47.8 best
42.5
MultilingualTasks in many languages.
—
50.9
InstructionTasks with strict prompt rules.
—
52.8
MathMath problems.
39.1
51.2 best

Benchmarks

11 shared
Only Claude Opus 4: 14Only GPT-OSS 120B: 37
Claude Opus 4
GPT-OSS 120B
72.7%112 / 220
75.8% best106 / 220
60.0%115 / 203
88.9% best66 / 203
83.1% best58 / 149
75.9%113 / 149
TaxEval v2Finance
71.9% best70 / 143
71.6%76 / 143
62.6%103 / 142
83.2% best49 / 142
71.7%90 / 135
78.5% best72 / 135
MMLU ProKnowledge
86.2% best45 / 135
79.2%98 / 135
AIMEMath
41.3%69 / 95
92.6% best18 / 95
MedQAHealthcare
92.9% best27 / 95
91.4%38 / 95
MGSMMultilingual
93.8% best8 / 74
92.0%24 / 74
90.4%24 / 59
94.8% best6 / 59