Is deepseek-r1:1.5b good for agentic (tool-using) tasks?

In our benchmarks, deepseek-r1:1.5b scores 16.8/100 for agentic workflows. It runs at about 128.0 tok/s, so if you want more speed, Tiel-Coder-35B-A3B-MLX-oQ4e-MTP is faster (~151.8 tok/s) and still scores well for agentic workflows (83.0/100).

How fast is deepseek-r1:1.5b for local inference?

Across 3 community benchmark runs, deepseek-r1:1.5b reaches up to 247.5 tok/s and averages 128.0 tok/s, with the fastest results on NVIDIA GeForce RTX 3060 Ti.

How much memory does deepseek-r1:1.5b need?

The leanest observed configuration used about 1.2 GB of memory (quantizations tested: Q4_K_M).

Which tools have been used to run deepseek-r1:1.5b?

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

Model benchmarks

deepseek-r1:1.5b local LLM performance

As of October 2026, deepseek-r1:1.5b runs at up to 247.5 tok/s for local inference (best of 3 community benchmark runs across 3 GPUs).

OllamaQ4_K_M

ShareReddit X

Model size

1.8B

Peak speed

247.5 tok/s

Average speed

128.0 tok/s

Avg PP

5946.6 tok/s

Min memory

1.2 GB

Max context

8,192 tokens

Avg output / run

8,511 tokens

Avg runtime / run

2m 5s

Avg quality

9.5

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 3 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	7.9	9.5	11.0
Agent Workflow	11.5	16.8	22.0
Code Generation	0.3	3.0	5.9
Role Play & Narrative	3.2	4.7	7.1
Research & Analysis	13.2	13.5	13.8

Performance by hardware and tool

Every hardware/tool/quantization combination deepseek-r1:1.5b has been benchmarked on, ranked by peak token generation speed. Last updated October 2026.

Hardware	Tool	Quant	Best tok/s	Avg tok/s	Memory	Context	Quality	Runs
NVIDIA GeForce RTX 3060 Ti	Ollama	Q4_K_M	247.5 tok/s	247.5 tok/s	1.3 GB	8,192 tokens	7.7	1
NVIDIA RTX 500 Ada Generation Laptop GPU	Ollama	Q4_K_M	82.2 tok/s	82.2 tok/s	1.2 GB	8,192 tokens	11.1	1
NVIDIA GeForce RTX 2050	Ollama	Q4_K_M	54.5 tok/s	54.5 tok/s	1.2 GB	8,192 tokens	9.7	1

Benchmark runs

All 3 deepseek-r1:1.5b runs submitted so far — expand one for its hardware, quality breakdown and per-scenario detail.

Open run page →

Frequently asked questions

Is deepseek-r1:1.5b good for agentic (tool-using) tasks?: In our benchmarks, deepseek-r1:1.5b scores 16.8/100 for agentic workflows. It runs at about 128.0 tok/s, so if you want more speed, Tiel-Coder-35B-A3B-MLX-oQ4e-MTP is faster (~151.8 tok/s) and still scores well for agentic workflows (83.0/100).
How fast is deepseek-r1:1.5b for local inference?: Across 3 community benchmark runs, deepseek-r1:1.5b reaches up to 247.5 tok/s and averages 128.0 tok/s, with the fastest results on NVIDIA GeForce RTX 3060 Ti.
How much memory does deepseek-r1:1.5b need?: The leanest observed configuration used about 1.2 GB of memory (quantizations tested: Q4_K_M).
Which tools have been used to run deepseek-r1:1.5b?: Benchmarks were submitted using Ollama. Results are community-contributed and updated as new runs arrive.