Claude 3.7 Sonnet Thinking vs GPT-OSS 20B

Change models
IndexAll categories, out of 100.
41.3 ±4.7
41.6 ±4.6 best
GapIndex points below the top.
−0.3 · no clear gap
Top
CoverageIndex weight that has results.
95% best
90%
SpeedOutput tokens per second.
—
350/s
EloLMArena rating and rank.
—
1288
ReleasedMonth of release.
2025-02
2025-08
Input / 1MUS dollars per 1M input tokens.
—
$0.075
Output / 1MUS dollars per 1M output tokens.
—
$0.30
1K chats2K tokens in, 500 out, each.
—
$0.30
1K documents20K tokens in, 1K out, each.
—
$1.80
1K agent tasks200K tokens in, 20K out, each.
—
$21.00
ContextMaximum tokens in one request.
200K best
131K
AgenticMulti-step tasks with tools.
39.0
39.8 best
CodingCode writing and repair.
39.4
48.7 best
ReasoningLogic problems and puzzles.
38.5 best
35.2
MultimodalTasks with images and text.
47.9
—
KnowledgeFacts and expert knowledge.
43.9 best
35.8
MultilingualTasks in many languages.
—
37.0
InstructionTasks with strict prompt rules.
42.4
49.8 best
MathMath problems.
36.4
45.7 best

Benchmarks

17 shared
Only Claude 3.7 Sonnet Thinking: 7Only GPT-OSS 20B: 20
Claude 3.7 Sonnet Thinking
GPT-OSS 20B
9.7%148 / 222
11.0% best141 / 222
0.9%124 / 214
1.4% best113 / 214
77.2% best107 / 199
68.8%134 / 199
82.5% best69 / 149
70.8%122 / 149
54.7%91 / 149
60.2% best86 / 149
48.3%82 / 146
65.1% best55 / 146
TaxEval v2Finance
74.0% best42 / 143
63.7%119 / 143
60.4%106 / 142
80.4% best69 / 142
78.2% best72 / 141
71.5%87 / 141
75.3% best81 / 135
68.9%97 / 135
MMLU ProKnowledge
82.7% best76 / 135
71.6%115 / 135
CorpFin v2Finance
60.4% best68 / 131
53.1%99 / 131
AIMEMath
44.6%64 / 95
86.0% best33 / 95
MedQAHealthcare
90.2% best47 / 95
82.9%65 / 95
MGSMMultilingual
93.0% best11 / 74
89.0%50 / 74
91.6%21 / 59
94.2% best10 / 59
61.3%21 / 27
92.1% best4 / 27