Is openai/gpt-oss-20b:2 good for agentic (tool-using) tasks?

In our benchmarks, openai/gpt-oss-20b:2 scores 87.2/100 for agentic workflows. It runs at about 154.8 tok/s, so if you want more speed, mudler/Ornith-1.5-35B-A3B-APEX-MTP-GGUF is faster (~190.9 tok/s) and still scores well for agentic workflows (85.7/100).

How fast is openai/gpt-oss-20b:2 for local inference?

Across 2 community benchmark runs, openai/gpt-oss-20b:2 reaches up to 158.7 tok/s and averages 154.8 tok/s, with the fastest results on NVIDIA GeForce RTX 5070.

Which tools have been used to run openai/gpt-oss-20b:2?

Benchmarks were submitted using LM Studio. Results are community-contributed and updated as new runs arrive.

Model benchmarks

openai/gpt-oss-20b:2 local LLM performance

As of August 2026, openai/gpt-oss-20b:2 runs at up to 158.7 tok/s for local inference (best of 2 community benchmark runs across 1 GPU).

LM Studio

ShareReddit X

Model size

20B

Peak speed

158.7 tok/s

Average speed

154.8 tok/s

Min memory

n/a

Max context

65,536 tokens

Avg runtime / run

42s

Avg quality

71.8

Benchmark runs

GPUs tested

Quality by task

Average LLM-judged quality (0–100) with the run-to-run spread shown as a P5–P95 band, overall and for each benchmark task, across all 2 runs. The low and high columns show how much the judge’s score varies between runs, and need at least two runs to display.

Task	P5 (low)	Avg	P95 (high)
Overall	70.7	71.8	73.0
Agent Workflow	85.5	87.2	88.9
Code Generation	56.6	60.7	64.8
Role Play & Narrative	78.3	78.4	78.5
Research & Analysis	59.8	61.0	62.1

Performance by hardware and tool

Every hardware/tool/quantization combination openai/gpt-oss-20b:2 has been benchmarked on, ranked by peak token generation speed. Last updated August 2026.

Hardware	Tool	Quant	Best tok/s	Avg tok/s	Memory	Context	Quality	Runs
NVIDIA GeForce RTX 5070	LM Studio	—	158.7 tok/s	154.8 tok/s	n/a	65,536 tokens	71.8	2

Benchmark runs

All 2 openai/gpt-oss-20b:2 runs submitted so far — expand one for its hardware, quality breakdown and per-scenario detail.

Open run page →

Frequently asked questions

Is openai/gpt-oss-20b:2 good for agentic (tool-using) tasks?: In our benchmarks, openai/gpt-oss-20b:2 scores 87.2/100 for agentic workflows. It runs at about 154.8 tok/s, so if you want more speed, mudler/Ornith-1.5-35B-A3B-APEX-MTP-GGUF is faster (~190.9 tok/s) and still scores well for agentic workflows (85.7/100).
How fast is openai/gpt-oss-20b:2 for local inference?: Across 2 community benchmark runs, openai/gpt-oss-20b:2 reaches up to 158.7 tok/s and averages 154.8 tok/s, with the fastest results on NVIDIA GeForce RTX 5070.
Which tools have been used to run openai/gpt-oss-20b:2?: Benchmarks were submitted using LM Studio. Results are community-contributed and updated as new runs arrive.