LLMs · Thinking Machines Lab

Inkling Small (276B-A12B)

Name: Inkling Small (276B-A12B)
Author: Thinking Machines Lab

Parameters: 266B (12.0B active)
Architecture: Mixture of experts
Max context: 1,024K tokens
License: apache-2.0
Released: 2026-07

How to run it

Quantized files · GGUF

Works in LM Studio, llama.cpp, Jan and Ollama. No sign-in needed.

huggingface.co/unsloth/Inkling-Small-GGUF (opens in a new tab)

ollama run hf.co/unsloth/Inkling-Small-GGUF

Original weights · Hugging Face

Full-precision files from Thinking Machines Lab, for fine-tuning, vLLM, Transformers.

huggingface.co/thinkingmachines/Inkling-Small (opens in a new tab)

No sign-in needed.

Links checked 1 day ago against Hugging Face, Ollama and GitHub.

Apps that run Inkling Small (276B-A12B)

Our pick for each kind of computer, assuming the model fits in its memory. Pick your own GPU or device in the Model Finder for an exact answer.

Mac (Apple silicon)

LM Studio

Runs the MLX build, typically 10–20% faster than GGUF for a model with 12B active parameters on Apple silicon.

Also good:

Windows PC, NVIDIA GPU

LM Studio

Point-and-click: search, download and chat in one window.

Also good:

Windows PC, AMD Radeon

Lemonade

AMD’s own app: llama.cpp tuned for Radeon on Vulkan and ROCm.

Also good:

Linux, NVIDIA GPU

LM Studio

Point-and-click: search, download and chat in one window.

Also good:

Linux server, 80 GB+ GPUs

vLLM

The full model fits in GPU memory at 8-bit, so it can batch many requests: best for serving a team or an app.

Also good:

App	Type	Opens it as	Systems	Engine	Start
LM Studio	Desktop app	GGUF or MLX	macOSWindowsLinux	llama.cpp, MLX	Open in LM Studio
Ollama	CLI + API	GGUF from Hugging Face	macOSWindowsLinux	llama.cpp, MLX	`ollama run hf.co/unsloth/Inkling-Small-GGUF:Q4_K_M`
llama.cpp	CLI + API	GGUF	macOSWindowsLinux	llama.cpp	`llama-server -hf unsloth/Inkling-Small-GGUF:Q4_K_M`
Jan	Desktop app	GGUF or MLX	macOSWindowsLinux	llama.cpp, MLX	Open in Jan
MLX LM	CLI + API	MLX	macOS	MLX	`mlx_lm.chat --model mlx-community/Inkling-Small-mlx-4bit`
vLLM	Server	Original weights	WSLWindowsLinux	vLLM	`vllm serve thinkingmachines/Inkling-Small`
SGLang	Server	Original weights	WSLWindowsLinux	SGLang	`python3 -m sglang.launch_server --model-path thinkingmachines/Inkling-Small`
Docker Model Runner	CLI + API	GGUF	macOSWindowsLinux	llama.cpp, vLLM	`docker model run hf.co/unsloth/Inkling-Small-GGUF:Q4_K_M`
Lemonade	Desktop app	GGUF	macOSWindowsLinux	llama.cpp, ONNX Runtime GenAI	`lemonade pull unsloth/Inkling-Small-GGUF:Q4_K_M`
Unsloth Desktop	Desktop app	GGUF or MLX	macOSWindowsLinux	llama.cpp, MLX, stable-diffusion.cpp	Open in Unsloth Desktop
Atomic Chat	Desktop app	GGUF or MLX	macOSWindowsLinux	llama.cpp, MLX-VLM	Open in Atomic Chat
Pi	Coding agent	GGUF or MLX	macOSWindowsLinux	uses llama.cpp or MLX LM	Point it at a llama.cpp or MLX LM server
Hermes Agent	Assistant	GGUF or MLX	macOSWSLWindowsLinux	uses llama.cpp or MLX LM	Point it at a llama.cpp or MLX LM server
OpenClaw	Assistant	GGUF or MLX	macOSWSLWindowsLinux	uses llama.cpp or MLX LM	Point it at a llama.cpp or MLX LM server
OpenCode	Coding agent	GGUF or MLX	macOSWindowsLinux	uses llama.cpp, Ollama or LM Studio	Point it at a llama.cpp or MLX LM server
Codex CLI	Coding agent	GGUF or MLX	macOSWSLWindowsLinux	uses Ollama or any OpenAI-compatible server	Point it at a llama.cpp or MLX LM server
OpenHands	Coding agent	GGUF or MLX	macOSWSLWindowsLinux	uses llama.cpp, Ollama or LM Studio	Point it at a llama.cpp or MLX LM server
Zed	Coding agent	GGUF or MLX	macOSWindowsLinux	uses llama.cpp, Ollama or LM Studio	Point it at a llama.cpp or MLX LM server
Cline	Coding agent	GGUF or MLX	macOSWindowsLinux	uses llama.cpp, Ollama or LM Studio	Point it at a llama.cpp or MLX LM server
Open Interpreter	Coding agent	GGUF or MLX	macOSWindowsLinux	uses llama.cpp, Ollama or LM Studio	Point it at a llama.cpp or MLX LM server
Goose	Coding agent	GGUF or MLX	macOSWindowsLinux	uses llama.cpp, Ollama or LM Studio	Point it at a llama.cpp or MLX LM server
Aider	Coding agent	GGUF or MLX	macOSWindowsLinux	uses llama.cpp, Ollama or LM Studio	Point it at a llama.cpp or MLX LM server
Crush	Coding agent	GGUF or MLX	macOSWindowsLinux	uses llama.cpp, Ollama or LM Studio	Point it at a llama.cpp or MLX LM server
Qwen Code	Coding agent	GGUF or MLX	macOSWindowsLinux	uses llama.cpp, Ollama or LM Studio	Point it at a llama.cpp or MLX LM server
Kilo Code	Coding agent	GGUF or MLX	macOSWindowsLinux	uses llama.cpp, Ollama or LM Studio	Point it at a llama.cpp or MLX LM server
Open WebUI	Assistant	GGUF or MLX	macOSWindowsLinux	uses llama.cpp, Ollama or LM Studio	Point it at a llama.cpp or MLX LM server
Khoj	Assistant	GGUF or MLX	macOSWindowsLinux	uses llama.cpp, Ollama or LM Studio	Point it at a llama.cpp or MLX LM server
Letta	Assistant	GGUF or MLX	macOSWindowsLinux	uses llama.cpp, Ollama or LM Studio	Point it at a llama.cpp or MLX LM server
KoboldCpp	Desktop app	GGUF	macOSWindowsLinux	llama.cpp	Get KoboldCpp
Text Generation WebUI	Desktop app	GGUF	macOSWindowsLinux	llama.cpp, Transformers, ExLlamaV3	Get Text Generation WebUI
llamafile	CLI + API	GGUF	macOSWindowsLinux	llama.cpp	Get llamafile
Msty Studio	Desktop app	GGUF or MLX	macOSWindowsLinux	Ollama, llama.cpp, MLX	Open in Msty Studio
Llama (macOS)	Desktop app	GGUF	macOS	llama.cpp	Open in Llama (macOS)
RecurseChat	Desktop app	GGUF	macOS	llama.cpp	Open in RecurseChat
Inferencer	Desktop app	MLX	macOS	MLX	Get Inferencer
node-llama-cpp	Library	GGUF	macOSWindowsLinux	llama.cpp	Get node-llama-cpp
exo	Server	MLX	macOSLinux	MLX	Get exo
Continue	Coding agent	GGUF or MLX	macOSWindowsLinux	uses Ollama, LM Studio or llama.cpp	Point it at a llama.cpp or MLX LM server
SillyTavern	Assistant	GGUF or MLX	macOSWindowsLinux	uses Ollama, LM Studio or llama.cpp	Point it at a llama.cpp or MLX LM server
LibreChat	Assistant	GGUF or MLX	macOSWindowsLinux	uses Ollama, LM Studio or llama.cpp	Point it at a llama.cpp or MLX LM server
Cherry Studio	Assistant	GGUF or MLX	macOSWindowsLinux	uses Ollama, LM Studio or llama.cpp	Point it at a llama.cpp or MLX LM server
Chatbox	Assistant	GGUF or MLX	macOSWindowsLinux	uses Ollama, LM Studio or llama.cpp	Point it at a llama.cpp or MLX LM server
PrivateGPT	Assistant	GGUF or MLX	macOSWindowsLinux	uses Ollama, LM Studio or llama.cpp	Point it at a llama.cpp or MLX LM server
PocketPal AI	Mobile app	GGUF	iOSAndroid	llama.cpp	App Store Google Play
ChatterUI	Mobile app	GGUF	Android	llama.cpp	Download the APK
LLM Farm	Mobile app	GGUF	iOS	llama.cpp	Not on the App Store or TestFlight for now (checked 2026-11-09).

Which apps can open a model depends on the files published for it: GGUF for llama.cpp-based apps, MLX for Apple’s engine, the original weights for vLLM and SGLang. Compare every app.

Memory needed

Weights plus KV cache and runtime overhead. Leave a little extra free for the operating system and other apps. Try another context length.

KV cache: sliding-window hybrid: 1 in 6 layers is global (window 512).

Quantization	Quality	Weights	Total at 8K context	Total at 32K context	Download
`Q2_K`	Low	106 GB	107 GB	108 GB	GGUF (UD-Q2_K_XL) · folder (opens in a new tab)
`Q3_K_M`	Low	130 GB	131 GB	132 GB	GGUF · folder (opens in a new tab)
`Q4_K_M`	Good	162 GB	163 GB	164 GB	GGUF · folder (opens in a new tab)
`Q5_K_M`	High	189 GB	190 GB	191 GB	GGUF · folder (opens in a new tab)
`Q6_K`	High	218 GB	219 GB	220 GB	GGUF · folder (opens in a new tab)
`Q8_0`	Very high	282 GB	283 GB	284 GB	GGUF · folder (opens in a new tab)
`F16`	Full	532 GB	533 GB	534 GB	GGUF · folder (opens in a new tab) Original weights (opens in a new tab)

Best hardware for Inkling Small (276B-A12B)

Quality	Hardware	Quant	Memory	Speed	Runs
92Excellent quality (A+) · 92 of 100: how capable the model is at Q6_K, whatever it runs on. Match score 96/100 (quality, fit, speed and memory use) sets the ranking.	Apple M3 Ultra (512 GB)	`Q6_K`	250 GB / 384 GB	42.4 tok/s	Runs great
89Excellent quality (A+) · 89 of 100: how capable the model is at Q4_K_M, whatever it runs on. Match score 94/100 (quality, fit, speed and memory use) sets the ranking.	Apple M3 Ultra (256 GB)	`Q4_K_M`	178 GB / 192 GB	58.0 tok/s	Runs welltight fit
78Great quality (A) · 78 of 100: how capable the model is at Q3_K_M, whatever it runs on. Match score 89/100 (quality, fit, speed and memory use) sets the ranking.	Apple M2 Ultra (192 GB)	`Q3_K_M`	143 GB / 144 GB	70.3 tok/s	Runs welltight fit
65Good quality (B) · 65 of 100: how capable the model is at Q2_K, whatever it runs on. Match score 69/100 (quality, fit, speed and memory use) sets the ranking.	NVIDIA RTX PRO 6000 Blackwell	`Q2_K`	112 GB / 96.0 GB	51.5 tok/s	Runs welloffloads to RAM
65Good quality (B) · 65 of 100: how capable the model is at Q2_K, whatever it runs on. Match score 68/100 (quality, fit, speed and memory use) sets the ranking.	NVIDIA GB10 (128 GB)	`Q2_K`	116 GB / 115 GB	34.1 tok/s	Runs wellover GPU memory limit
93Excellent quality (A+) · 93 of 100: how capable the model is at Q8_0, whatever it runs on. Match score 97/100 (quality, fit, speed and memory use) sets the ranking.	Apple M5 Ultra (512 GB)	`Q8_0`	314 GB / 384 GB	44.7 tok/s	Runs great
92Excellent quality (A+) · 92 of 100: how capable the model is at Q6_K, whatever it runs on. Match score 96/100 (quality, fit, speed and memory use) sets the ranking.	AMD Instinct MI350X	`Q6_K`	233 GB / 288 GB	177 tok/s	Runs great
90Excellent quality (A+) · 90 of 100: how capable the model is at Q5_K_M, whatever it runs on. Match score 95/100 (quality, fit, speed and memory use) sets the ranking.	AMD Instinct MI325X	`Q5_K_M`	203 GB / 256 GB	173 tok/s	Runs great
90Excellent quality (A+) · 90 of 100: how capable the model is at Q5_K_M, whatever it runs on. Match score 95/100 (quality, fit, speed and memory use) sets the ranking.	NVIDIA GB300 Grace Blackwell Ultra Desktop Superchip	`Q5_K_M`	202 GB / 252 GB	194 tok/s	Runs great
89Excellent quality (A+) · 89 of 100: how capable the model is at Q4_K_M, whatever it runs on. Match score 94/100 (quality, fit, speed and memory use) sets the ranking.	AMD Instinct MI300X	`Q4_K_M`	173 GB / 192 GB	182 tok/s	Runs welltight fit

Check it on your own hardware

Run it on a cluster

Devices that can’t run Inkling Small (276B-A12B) alone can run it together. These example clusters do:

Two gaming PCs (RTX 3090 + RTX 4090)14.9 tok/s at Q2_K with llama.cpp RPC

Try it with your own devices

Estimates come from published hardware specs and our open formulas, not from running the model on your machine. Real speed varies with runtime, drivers and settings. Numbers marked “Measured” come from a linked benchmark.