Is Swift-1.5-Qwen3.8-27B-GSQ-RCO-IQ3_XXS-mtp good for coding?

In our benchmarks, Swift-1.5-Qwen3.8-27B-GSQ-RCO-IQ3_XXS-mtp scores 73.6/100 for coding. It runs at about 44.9 tok/s, so if you want more speed, ornith-1.5-35b-a3b-apex-i-quality is faster (~73.2 tok/s) and still scores well for coding (83.1/100).

Is Swift-1.5-Qwen3.8-27B-GSQ-RCO-IQ3_XXS-mtp good for agentic (tool-using) tasks?

In our benchmarks, Swift-1.5-Qwen3.8-27B-GSQ-RCO-IQ3_XXS-mtp scores 84.7/100 for agentic workflows. It runs at about 44.9 tok/s, so if you want more speed, Nex-N2.5-mini-APEX-Mini is faster (~88.2 tok/s) and still scores well for agentic workflows (91.3/100).

How fast is Swift-1.5-Qwen3.8-27B-GSQ-RCO-IQ3_XXS-mtp for local inference?

Across 20 community benchmark runs, Swift-1.5-Qwen3.8-27B-GSQ-RCO-IQ3_XXS-mtp reaches up to 52.8 tok/s and averages 44.9 tok/s, with the fastest results on AMD Radeon RX 7800 XT.

How much memory does Swift-1.5-Qwen3.8-27B-GSQ-RCO-IQ3_XXS-mtp need?

The leanest observed configuration used about 14.0 GB of memory.

Model benchmarks

Swift-1.5-Qwen3.8-27B-GSQ-RCO-IQ3_XXS-mtp local LLM performance

As of November 2026, Swift-1.5-Qwen3.8-27B-GSQ-RCO-IQ3_XXS-mtp runs at up to 52.8 tok/s for local inference (best of 20 community benchmark runs across 3 GPUs).

llama.cppllama.cpp k4 v4 2048 budgetllama.cpp k4 v4 4096 budgetllama.cpp k4 v4 8192 budgetllama.cpp k8 v4 12288 budgetllama.cpp k8 v4 16384 budgetllama.cpp q8 q8llama.cpp q8 q8 -1 budget

ShareReddit X

Model size

27B

Peak speed

52.8 tok/s

Average speed

44.9 tok/s

Avg PP

347.9 tok/s

Min memory

14.0 GB

Max context

131,072 tokens

Avg output / run

38,134 tokens

Avg runtime / run

14m 27s

Avg quality

84.0

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 20 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	77.4	84.0	87.4
Agent Workflow	75.8	84.7	91.0
Code Generation	52.8	73.6	85.4
Role Play & Narrative	90.6	92.6	95.0
Research & Analysis	80.0	84.9	89.4

Performance by hardware and tool

Every hardware/tool/quantization combination Swift-1.5-Qwen3.8-27B-GSQ-RCO-IQ3_XXS-mtp has been benchmarked on, ranked by peak token generation speed. Last updated November 2026.

Hardware	Tool	Quant	Best tok/s	Avg tok/s	Memory	Context	Quality	Runs
AMD Radeon RX 7800 XT	llama.cpp	—	52.8 tok/s	46.1 tok/s	14.0 GB	131,072 tokens	85.3	9
AMD Radeon RX 7800 XT	llama.cpp q8 q8 -1 budget	—	50.6 tok/s	50.6 tok/s	17.2 GB	128,000 tokens	86.6	1
AMD Radeon RX 7800 XT	llama.cpp q8 q8	—	49.5 tok/s	49.5 tok/s	17.2 GB	128,000 tokens	80.8	1
AMD Radeon RX 7800 XT	llama.cpp k4 v4 8192 budget	—	49.4 tok/s	49.4 tok/s	15.2 GB	128,000 tokens	85.4	1
AMD Radeon RX 7800 XT	llama.cpp k4 v4 2048 budget	—	47.8 tok/s	47.8 tok/s	15.2 GB	128,000 tokens	80.0	1
AMD Radeon RX 7800 XT	llama.cpp k4 v4 4096 budget	—	47.7 tok/s	47.7 tok/s	15.2 GB	128,000 tokens	86.3	1
AMD Radeon RX 7800 XT	llama.cpp k8 v4 12288 budget	—	47.4 tok/s	47.4 tok/s	16.2 GB	128,000 tokens	87.4	1
AMD Radeon RX 7800 XT	llama.cpp k8 v4 16384 budget	—	45.0 tok/s	45.0 tok/s	16.2 GB	128,000 tokens	84.0	1
NVIDIA GeForce RTX 5060	llama.cpp	—	39.7 tok/s	39.2 tok/s	16.2 GB	131,072 tokens	79.8	3
AMD Radeon RX 6700/6700 XT/6750 XT / 6800M/6850M XT	llama.cpp	—	28.6 tok/s	28.6 tok/s	n/a	32,768 tokens	81.4	1

Benchmark runs

All 20 Swift-1.5-Qwen3.8-27B-GSQ-RCO-IQ3_XXS-mtp runs submitted so far — expand one for its hardware, quality breakdown and per-scenario detail.

Open run page →

Frequently asked questions

Is Swift-1.5-Qwen3.8-27B-GSQ-RCO-IQ3_XXS-mtp good for coding?: In our benchmarks, Swift-1.5-Qwen3.8-27B-GSQ-RCO-IQ3_XXS-mtp scores 73.6/100 for coding. It runs at about 44.9 tok/s, so if you want more speed, ornith-1.5-35b-a3b-apex-i-quality is faster (~73.2 tok/s) and still scores well for coding (83.1/100).
Is Swift-1.5-Qwen3.8-27B-GSQ-RCO-IQ3_XXS-mtp good for agentic (tool-using) tasks?: In our benchmarks, Swift-1.5-Qwen3.8-27B-GSQ-RCO-IQ3_XXS-mtp scores 84.7/100 for agentic workflows. It runs at about 44.9 tok/s, so if you want more speed, Nex-N2.5-mini-APEX-Mini is faster (~88.2 tok/s) and still scores well for agentic workflows (91.3/100).
How fast is Swift-1.5-Qwen3.8-27B-GSQ-RCO-IQ3_XXS-mtp for local inference?: Across 20 community benchmark runs, Swift-1.5-Qwen3.8-27B-GSQ-RCO-IQ3_XXS-mtp reaches up to 52.8 tok/s and averages 44.9 tok/s, with the fastest results on AMD Radeon RX 7800 XT.
How much memory does Swift-1.5-Qwen3.8-27B-GSQ-RCO-IQ3_XXS-mtp need?: The leanest observed configuration used about 14.0 GB of memory.
Which tools have been used to run Swift-1.5-Qwen3.8-27B-GSQ-RCO-IQ3_XXS-mtp?: Benchmarks were submitted using llama.cpp, llama.cpp k4 v4 2048 budget, llama.cpp k4 v4 4096 budget, llama.cpp k4 v4 8192 budget, llama.cpp k8 v4 12288 budget, llama.cpp k8 v4 16384 budget, llama.cpp q8 q8, llama.cpp q8 q8 -1 budget. Results are community-contributed and updated as new runs arrive.