GPT-5 (high) vs GPT-OSS 20B

Change models
IndexAll categories, out of 100.
53.6 ±5.0 best
41.7 ±4.6
GapIndex points below the top.
Top
−11.9
CoverageIndex weight that has results.
95% best
90%
SpeedOutput tokens per second.
76/s
310/s best
EloLMArena rating and rank.
1406 best
1288
ReleasedMonth of release.
2025-08
2025-08
Input / 1MUS dollars per 1M input tokens.
$1.25
$0.03 best
Output / 1MUS dollars per 1M output tokens.
$10.00
$0.13 best
1K chats2K tokens in, 500 out, each.
$7.50
$0.125 best
1K documents20K tokens in, 1K out, each.
$35.00
$0.73 best
1K agent tasks200K tokens in, 20K out, each.
$450.00
$8.60 best
ContextMaximum tokens in one request.
400K best
131K
AgenticMulti-step tasks with tools.
49.7 best
39.8
CodingCode writing and repair.
59.5 best
48.9
ReasoningLogic problems and puzzles.
54.4 best
35.3
MultimodalTasks with images and text.
51.7
—
KnowledgeFacts and expert knowledge.
56.1 best
36.0
MultilingualTasks in many languages.
—
37.0
InstructionTasks with strict prompt rules.
57.3 best
49.9
MathMath problems.
49.2 best
45.9

Benchmarks

7 shared
Only GPT-5 (high): 8Only GPT-OSS 20B: 30
GPT-5 (high)
GPT-OSS 20B
28.5% best87 / 222
11.0%141 / 222
78.2% best68 / 217
34.7%175 / 217
5.7% best78 / 214
1.4%113 / 214
85.4% best76 / 199
68.8%134 / 199
84.8% best58 / 149
60.2%86 / 149
73.1% best32 / 146
65.1%55 / 146
20.9% best81 / 133
0.0%101 / 133