Model benchmarks

thinkingmachines/inkling-2-mini local LLM performance

As of November 2026, thinkingmachines/inkling-2-mini runs at up to 79.3 tok/s for local inference (best of 4 community benchmark runs across 2 GPUs).

LM StudioOpenAI-compatibleQ4_K_M
ShareRedditX

Model size

26B

Peak speed

79.3 tok/s

Average speed

50.7 tok/s

Avg PP

730.5 tok/s

Min memory

11.3 GB

Max context

131,072 tokens

Avg output / run

31,037 tokens

Avg runtime / run

10m 38s

Avg quality

83.2

Benchmark runs

4

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 4 runs. The low and high columns show how much the judge’s score varies between runs, and need at least two runs to display.

TaskP5 (low)AvgP95 (high)
Overall79.883.287.2
Agent Workflow89.691.292.7
Code Generation81.383.786.4
Role Play & Narrative56.571.687.1
Research & Analysis84.286.488.3

Performance by hardware and tool

Every hardware/tool/quantization combination thinkingmachines/inkling-2-mini has been benchmarked on, ranked by peak token generation speed. Last updated November 2026.

HardwareToolQuantBest tok/sAvg tok/sMemoryContextQualityRuns
NVIDIA GeForce RTX 3080 TiOpenAI-compatible—79.3 tok/s79.3 tok/s11.3 GB65,536 tokens79.71
Apple M5 MaxLM StudioQ4_K_M41.3 tok/s41.2 tok/s20.0 GB131,072 tokens84.43

Benchmark runs

All 4 thinkingmachines/inkling-2-mini runs submitted so far — expand one for its hardware, quality breakdown and per-scenario detail.

Frequently asked questions

Is thinkingmachines/inkling-2-mini good for coding?
In our benchmarks, thinkingmachines/inkling-2-mini scores 83.7/100 for coding. It runs at about 50.7 tok/s, so if you want more speed, ornith-1.5-35b-a3b-apex-mtp-i-quality is faster (~105.2 tok/s) and still scores well for coding (83.6/100).
Is thinkingmachines/inkling-2-mini good for agentic (tool-using) tasks?
In our benchmarks, thinkingmachines/inkling-2-mini scores 91.2/100 for agentic workflows, among the top 3 for agentic workflows on consumer hardware (≤24 GB VRAM). It runs at about 50.7 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 thinkingmachines/inkling-2-mini for local inference?
Across 4 community benchmark runs, thinkingmachines/inkling-2-mini reaches up to 79.3 tok/s and averages 50.7 tok/s, with the fastest results on NVIDIA GeForce RTX 3080 Ti.
How much memory does thinkingmachines/inkling-2-mini need?
The leanest observed configuration used about 11.3 GB of memory (quantizations tested: Q4_K_M).
Which tools have been used to run thinkingmachines/inkling-2-mini?
Benchmarks were submitted using LM Studio, OpenAI-compatible. Results are community-contributed and updated as new runs arrive.