PipelineScore
← Back to leaderboard
Head-to-head

Llama 4 70B Instruct vs openai/gpt-oss-20b

openai/gpt-oss-20b wins+10.9 pts (+13%)on the balanced composite.
meta
Llama 4 70B Instruct
81.1MAINLINE
local
Winner
openai/gpt-oss-20b
92.0TRUNK

Category-by-category

Higher bar wins each row.

Llama 4 70B Instruct
Category
openai/gpt-oss-20b
-16.883.3
Code
100.0+16.8
+5.880.8
Reason
75.0-5.8
-7.380.3
Tool Use
87.5+7.3
-15.085.0
RAG
100.0+15.0
-29.569.8
Speed
99.3+29.5
Llama 4 70B Instruct
Provider
meta
Released
2026-01-20
Context
256K
Lab
Community
openai/gpt-oss-20b
Provider
local
Released
Context
0K
Lab
Community