Model benchmarks
Tiel-Coder-35B-A3B-MLX-oQ6e-MTP local LLM performance
As of October 2026, Tiel-Coder-35B-A3B-MLX-oQ6e-MTP runs at up to 214.4 tok/s for local inference (best of 10 community benchmark runs across 2 GPUs).
Model size
35B
Peak speed
214.4 tok/s
Average speed
145.8 tok/s
Avg PP
3906.4 tok/s
Min memory
27.2 GB
Max context
262,144 tokens
Avg output / run
25,156 tokens
Avg runtime / run
3m 1s
Avg quality
81.2
Benchmark runs
10
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 10 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 | 75.3 | 81.2 | 86.5 |
| Agent Workflow | 77.3 | 82.5 | 88.1 |
| Code Generation | 54.4 | 75.7 | 83.6 |
| Role Play & Narrative | 76.5 | 82.0 | 88.6 |
| Research & Analysis | 80.4 | 84.6 | 88.2 |
Performance by hardware and tool
Every hardware/tool/quantization combination Tiel-Coder-35B-A3B-MLX-oQ6e-MTP 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 |
|---|---|---|---|---|---|---|---|---|
| Apple M5 Ultra | oMLX | oQ6e | 214.4 tok/s | 214.4 tok/s | 32.6 GB | 65,536 tokens | 79.8 | 1 |
| Apple M5 Max | oMLX | oQ6e | 174.4 tok/s | 138.1 tok/s | 27.2 GB | 262,144 tokens | 81.3 | 9 |
Benchmark runs
All 10 Tiel-Coder-35B-A3B-MLX-oQ6e-MTP runs submitted so far — expand one for its hardware, quality breakdown and per-scenario detail.
Frequently asked questions
- Is Tiel-Coder-35B-A3B-MLX-oQ6e-MTP good for agentic (tool-using) tasks?
- In our benchmarks, Tiel-Coder-35B-A3B-MLX-oQ6e-MTP scores 82.5/100 for agentic workflows. It runs at about 145.8 tok/s, so if you want more speed, swift-1.5-qwen3.8-flash-next-iq3_s is faster (~259.0 tok/s) and still scores well for agentic workflows (88.9/100).
- How fast is Tiel-Coder-35B-A3B-MLX-oQ6e-MTP for local inference?
- Across 10 community benchmark runs, Tiel-Coder-35B-A3B-MLX-oQ6e-MTP reaches up to 214.4 tok/s and averages 145.8 tok/s, with the fastest results on Apple M5 Ultra.
- How much memory does Tiel-Coder-35B-A3B-MLX-oQ6e-MTP need?
- The leanest observed configuration used about 27.2 GB of memory (quantizations tested: oQ6e).
- Which tools have been used to run Tiel-Coder-35B-A3B-MLX-oQ6e-MTP?
- Benchmarks were submitted using oMLX. Results are community-contributed and updated as new runs arrive.