Is Cyber-Tiel-Coder-35B-A3B-MTP-UD-IQ3_XXS good for coding?

In our benchmarks, Cyber-Tiel-Coder-35B-A3B-MTP-UD-IQ3_XXS scores 76.6/100 for coding. It runs at about 66.2 tok/s, so if you want more speed, Nex-N2.5-mini-IQ3_XXS is faster (~91.4 tok/s) and still scores well for coding (81.6/100).

Is Cyber-Tiel-Coder-35B-A3B-MTP-UD-IQ3_XXS good for agentic (tool-using) tasks?

In our benchmarks, Cyber-Tiel-Coder-35B-A3B-MTP-UD-IQ3_XXS scores 81.9/100 for agentic workflows. It runs at about 66.2 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 Cyber-Tiel-Coder-35B-A3B-MTP-UD-IQ3_XXS for local inference?

Across 46 community benchmark runs, Cyber-Tiel-Coder-35B-A3B-MTP-UD-IQ3_XXS reaches up to 119.3 tok/s and averages 66.2 tok/s, with the fastest results on NVIDIA GeForce RTX 5060.

How much memory does Cyber-Tiel-Coder-35B-A3B-MTP-UD-IQ3_XXS need?

The leanest observed configuration used about 19.1 GB of memory.

Which tools have been used to run Cyber-Tiel-Coder-35B-A3B-MTP-UD-IQ3_XXS?

Benchmarks were submitted using LM Studio, llama.cpp. Results are community-contributed and updated as new runs arrive.

Model benchmarks

Cyber-Tiel-Coder-35B-A3B-MTP-UD-IQ3_XXS local LLM performance

As of October 2026, Cyber-Tiel-Coder-35B-A3B-MTP-UD-IQ3_XXS runs at up to 119.3 tok/s for local inference (best of 46 community benchmark runs across 3 GPUs).

LM Studiollama.cpp

ShareReddit X

Model size

35B

Peak speed

119.3 tok/s

Average speed

66.2 tok/s

Avg PP

512.9 tok/s

Min memory

19.1 GB

Max context

131,072 tokens

Avg output / run

20,853 tokens

Avg runtime / run

5m 19s

Avg quality

80.8

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 46 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	69.9	80.8	86.3
Agent Workflow	51.0	81.9	91.0
Code Generation	63.2	76.6	85.7
Role Play & Narrative	69.2	82.6	90.4
Research & Analysis	71.4	82.2	88.8

Performance by hardware and tool

Every hardware/tool/quantization combination Cyber-Tiel-Coder-35B-A3B-MTP-UD-IQ3_XXS 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
NVIDIA GeForce RTX 5060	llama.cpp	—	119.3 tok/s	116.1 tok/s	n/a	131,072 tokens	78.8	4
AMD Radeon RX 7900 XTX	LM Studio	—	113.8 tok/s	113.8 tok/s	19.1 GB	65,536 tokens	85.8	1
Tesla P100-PCIE-16GB	llama.cpp	—	68.5 tok/s	60.2 tok/s	n/a	131,072 tokens	80.9	41

Benchmark runs

All 46 Cyber-Tiel-Coder-35B-A3B-MTP-UD-IQ3_XXS runs submitted so far — expand one for its hardware, quality breakdown and per-scenario detail.

Open run page →

Frequently asked questions

Is Cyber-Tiel-Coder-35B-A3B-MTP-UD-IQ3_XXS good for coding?: In our benchmarks, Cyber-Tiel-Coder-35B-A3B-MTP-UD-IQ3_XXS scores 76.6/100 for coding. It runs at about 66.2 tok/s, so if you want more speed, Nex-N2.5-mini-IQ3_XXS is faster (~91.4 tok/s) and still scores well for coding (81.6/100).
Is Cyber-Tiel-Coder-35B-A3B-MTP-UD-IQ3_XXS good for agentic (tool-using) tasks?: In our benchmarks, Cyber-Tiel-Coder-35B-A3B-MTP-UD-IQ3_XXS scores 81.9/100 for agentic workflows. It runs at about 66.2 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 Cyber-Tiel-Coder-35B-A3B-MTP-UD-IQ3_XXS for local inference?: Across 46 community benchmark runs, Cyber-Tiel-Coder-35B-A3B-MTP-UD-IQ3_XXS reaches up to 119.3 tok/s and averages 66.2 tok/s, with the fastest results on NVIDIA GeForce RTX 5060.
How much memory does Cyber-Tiel-Coder-35B-A3B-MTP-UD-IQ3_XXS need?: The leanest observed configuration used about 19.1 GB of memory.
Which tools have been used to run Cyber-Tiel-Coder-35B-A3B-MTP-UD-IQ3_XXS?: Benchmarks were submitted using LM Studio, llama.cpp. Results are community-contributed and updated as new runs arrive.