Model benchmarks
30cmtonyfaker/Qwen3.8-Flash-Next-REAP-288-MLX-Serve-4bit-MTP local LLM performance
As of September 2026, 30cmtonyfaker/Qwen3.8-Flash-Next-REAP-288-MLX-Serve-4bit-MTP runs at up to 81.4 tok/s for local inference (best of 3 community benchmark runs across 1 GPU).
Model size
Unknown
Peak speed
81.4 tok/s
Average speed
77.4 tok/s
Avg PP
922.8 tok/s
Min memory
n/a
Max context
262.144 tokens
Avg output / run
4.984 tokens
Avg runtime / run
1m 8s
Avg quality
54.0
Benchmark runs
3
GPUs tested
1
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 | 51.2 | 54.0 | 57.6 |
| Agent Workflow | 47.8 | 61.5 | 74.1 |
| Code Generation | 0.0 | 7.6 | 20.5 |
| Role Play & Narrative | 76.4 | 78.8 | 82.4 |
| Research & Analysis | 64.5 | 68.2 | 73.2 |
Performance by hardware and tool
Every hardware/tool/quantization combination 30cmtonyfaker/Qwen3.8-Flash-Next-REAP-288-MLX-Serve-4bit-MTP has been benchmarked on, ranked by peak token generation speed. Last updated September 2026.
| Hardware | Tool | Quant | Best tok/s | Avg tok/s | Memory | Context | Quality | Runs |
|---|---|---|---|---|---|---|---|---|
| Apple M5 Max | vLLM | — | 81.4 tok/s | 77.4 tok/s | n/a | 262.144 tokens | 54.0 | 3 |
Benchmark runs
All 3 30cmtonyfaker/Qwen3.8-Flash-Next-REAP-288-MLX-Serve-4bit-MTP runs submitted so far — expand one for its hardware, quality breakdown and per-scenario detail.
Frequently asked questions
- Is 30cmtonyfaker/Qwen3.8-Flash-Next-REAP-288-MLX-Serve-4bit-MTP good for coding?
- In our benchmarks, 30cmtonyfaker/Qwen3.8-Flash-Next-REAP-288-MLX-Serve-4bit-MTP scores 7.6/100 for coding. It runs at about 77.4 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 30cmtonyfaker/Qwen3.8-Flash-Next-REAP-288-MLX-Serve-4bit-MTP good for agentic (tool-using) tasks?
- In our benchmarks, 30cmtonyfaker/Qwen3.8-Flash-Next-REAP-288-MLX-Serve-4bit-MTP scores 61.5/100 for agentic workflows. It runs at about 77.4 tok/s, so if you want more speed, Ornith-1.5-35B-A3B-BigBang-oQ8e-mtp is faster (~102.7 tok/s) and still scores well for agentic workflows (91.1/100).
- How fast is 30cmtonyfaker/Qwen3.8-Flash-Next-REAP-288-MLX-Serve-4bit-MTP for local inference?
- Across 3 community benchmark runs, 30cmtonyfaker/Qwen3.8-Flash-Next-REAP-288-MLX-Serve-4bit-MTP reaches up to 81.4 tok/s and averages 77.4 tok/s, with the fastest results on Apple M5 Max.
- How much memory does 30cmtonyfaker/Qwen3.8-Flash-Next-REAP-288-MLX-Serve-4bit-MTP need?
- The leanest observed configuration used about n/a of memory.
- Which tools have been used to run 30cmtonyfaker/Qwen3.8-Flash-Next-REAP-288-MLX-Serve-4bit-MTP?
- Benchmarks were submitted using vLLM. Results are community-contributed and updated as new runs arrive.