Model benchmarks
Tiel-Coder-35B-A3B-MTP-UD-Q4_K_XL local LLM performance
As of October 2026, Tiel-Coder-35B-A3B-MTP-UD-Q4_K_XL runs at up to 53.3 tok/s for local inference (best of 6 community benchmark runs across 4 GPUs).
Model size
35B
Peak speed
53.3 tok/s
Average speed
38.3 tok/s
Avg PP
458.6 tok/s
Min memory
12.9 GB
Max context
262,144 tokens
Avg output / run
16,247 tokens
Avg runtime / run
9m 10s
Avg quality
73.3
Benchmark runs
6
GPUs tested
4
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 6 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 | 56.9 | 73.3 | 84.0 |
| Agent Workflow | 59.3 | 78.7 | 93.1 |
| Code Generation | 67.6 | 75.8 | 83.9 |
| Role Play & Narrative | 23.2 | 61.7 | 90.1 |
| Research & Analysis | 66.7 | 76.9 | 88.1 |
Performance by hardware and tool
Every hardware/tool/quantization combination Tiel-Coder-35B-A3B-MTP-UD-Q4_K_XL 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 |
|---|---|---|---|---|---|---|---|---|
| AMD Radeon 8050S | llama.cpp | Q4_K | 53.3 tok/s | 53.3 tok/s | n/a | 65,536 tokens | 82.7 | 1 |
| Tesla P100-PCIE-16GB | llama.cpp | Q4_K | 38.8 tok/s | 38.5 tok/s | n/a | 65,536 tokens | 63.4 | 3 |
| AMD Radeon RX 6700/6700 XT/6750 XT / 6800M/6850M XT | llama.cpp | Q4_K | 33.9 tok/s | 33.9 tok/s | n/a | 262,144 tokens | 82.4 | 1 |
| AMD Radeon RX 9060 XT | Ollama | Q4_K_M | 27.0 tok/s | 27.0 tok/s | 12.9 GB | 65,536 tokens | 84.5 | 1 |
Benchmark runs
All 6 Tiel-Coder-35B-A3B-MTP-UD-Q4_K_XL runs submitted so far — expand one for its hardware, quality breakdown and per-scenario detail.
Frequently asked questions
- Is Tiel-Coder-35B-A3B-MTP-UD-Q4_K_XL good for coding?
- In our benchmarks, Tiel-Coder-35B-A3B-MTP-UD-Q4_K_XL scores 75.8/100 for coding. It runs at about 38.3 tok/s, so if you want more speed, nvidia/Qwen3.8-27B-NVFP4 is faster (~63.3 tok/s) and still scores well for coding (82.7/100).
- Is Tiel-Coder-35B-A3B-MTP-UD-Q4_K_XL good for agentic (tool-using) tasks?
- In our benchmarks, Tiel-Coder-35B-A3B-MTP-UD-Q4_K_XL scores 78.7/100 for agentic workflows. It runs at about 38.3 tok/s, so if you want more speed, swift-iq2_xs is faster (~91.5 tok/s) and still scores well for agentic workflows (91.4/100).
- How fast is Tiel-Coder-35B-A3B-MTP-UD-Q4_K_XL for local inference?
- Across 6 community benchmark runs, Tiel-Coder-35B-A3B-MTP-UD-Q4_K_XL reaches up to 53.3 tok/s and averages 38.3 tok/s, with the fastest results on AMD Radeon 8050S.
- How much memory does Tiel-Coder-35B-A3B-MTP-UD-Q4_K_XL need?
- The leanest observed configuration used about 12.9 GB of memory (quantizations tested: Q4_K, Q4_K_M).
- Which tools have been used to run Tiel-Coder-35B-A3B-MTP-UD-Q4_K_XL?
- Benchmarks were submitted using Ollama, llama.cpp. Results are community-contributed and updated as new runs arrive.