Model benchmarks
qwen3.6-14b-a3b-fablevibes local LLM performance
As of August 2026, qwen3.6-14b-a3b-fablevibes runs at up to 95.3 tok/s for local inference (best of 3 community benchmark runs across 2 GPUs).
Model size
14B
Peak speed
95.3 tok/s
Average speed
71.4 tok/s
Min memory
8.7 GB
Max context
16.384 tokens
Avg quality
50.5
Benchmark runs
3
GPUs tested
2
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 | 45.1 | 50.5 | 57.6 |
| Agent Workflow | 57.9 | 62.2 | 66.6 |
| Code Generation | 32.4 | 39.0 | 44.2 |
| Role Play & Narrative | 24.4 | 40.2 | 61.1 |
| Research & Analysis | 53.1 | 60.7 | 70.6 |
Performance by hardware and tool
Every hardware/tool/quantization combination qwen3.6-14b-a3b-fablevibes 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 |
|---|---|---|---|---|---|---|---|---|
| AMD Radeon RX 6800 XT | LM Studio | — | 95.3 tok/s | 84.4 tok/s | 8.7 GB | 8.192 tokens | 46.5 | 2 |
| NVIDIA GeForce RTX 4070 Ti SUPER | LM Studio | — | 45.2 tok/s | 45.2 tok/s | 14.5 GB | 16.384 tokens | 58.6 | 1 |
Benchmark runs
All 3 qwen3.6-14b-a3b-fablevibes runs submitted so far — expand one for its hardware, quality breakdown and per-scenario detail.
Frequently asked questions
- Is qwen3.6-14b-a3b-fablevibes good for coding?
- In our benchmarks, qwen3.6-14b-a3b-fablevibes scores 39.0/100 for coding. It runs at about 71.4 tok/s, so if you want more speed, unsloth/Qwen3.8-27B-GGUF:IQ3_S is faster (~104.3 tok/s) and still scores well for coding (83.0/100).
- Is qwen3.6-14b-a3b-fablevibes good for agentic (tool-using) tasks?
- In our benchmarks, qwen3.6-14b-a3b-fablevibes scores 62.2/100 for agentic workflows. It runs at about 71.4 tok/s, so if you want more speed, Tiel-Coder-35B-A3B-Q4_K_S is faster (~165.0 tok/s) and still scores well for agentic workflows (89.3/100).
- How fast is qwen3.6-14b-a3b-fablevibes for local inference?
- Across 3 community benchmark runs, qwen3.6-14b-a3b-fablevibes reaches up to 95.3 tok/s and averages 71.4 tok/s, with the fastest results on AMD Radeon RX 6800 XT.
- How much memory does qwen3.6-14b-a3b-fablevibes need?
- The leanest observed configuration used about 8.7 GB of memory.
- Which tools have been used to run qwen3.6-14b-a3b-fablevibes?
- Benchmarks were submitted using LM Studio. Results are community-contributed and updated as new runs arrive.