Skip to content

Inkling 2 Mini VRAM Requirements

Thinking Machines Lab, United States · Released November 20, 2026New

Inkling 2 Mini has 26.4B parameters (3.4B active per token). In BF16 its weights alone take 49.2 GiB; quantized to 4-bit, about 15.8 GiB. Each 32K-token request adds 1.16 GiB of KV cache. Cheapest way to run it today: 1 × GeForce RTX 3090 (24 GB) at $0.109 per hour. It reads text and images, calls tools and lets you set its reasoning effort.

Size your own setup in the calculator →

Inkling 2 Mini at a glance

Made by
Thinking Machines Lab, United States
Released
November 20, 2026
Parameters
26.4B, 3.4B active
Context window
1M tokens
License
Apache 2.0, commercial use allowed
Reads
Text and images
Tool calling
Supported
Reasoning
Adjustable reasoning effort
Runs with
vLLM, llama.cpp (Ollama, LM Studio), MLX (Mac) · vLLM recipe (opens in a new tab)
On Ollama
ollama run inkling-2-mini · safetensors, 17.6 GiB 20K pulls (opens in a new tab)

VRAM by precision and context length

Weights plus the KV cache of one request (BF16 cache) and the model's fixed recurrent state, in GiB. Add 1 to 3 GiB for the inference engine itself.

Weights formatWeights only+ 8K context+ 32K context+ 128K context+ 1M context
NVFP4 / MXFP8 experts + BF16 (native)18.218.619.322.350.3
BF16 / FP1649.249.650.353.381.3
FP8 / INT825.425.826.629.657.6
4-bit (GGUF Q4_K_M)15.816.317.020.048.0

Inkling 2 Mini GGUF files

Exact size of each quantization in AtomicChat/Inkling-2-Mini-GGUF (opens in a new tab), and the smallest GPU that runs it with llama.cpp, Ollama or LM Studio and an 8K-token context.

QuantizationFile sizeSmallest GPU
BF1648.6 GiBRTX PRO 6000 Blackwell (96 GB)
Q8_025.8 GiBRTX 5090 (32 GB)
Q6_K21.8 GiBRTX 5090 (32 GB)
Q5_K_M18.5 GiBRTX 4090 (24 GB)
Q4_K_M16.2 GiBRTX 4090 (24 GB)
IQ4_XS14.0 GiBRTX 4090 (24 GB)
IQ1_S6.41 GiBRTX 4070 / 4070 SUPER (12 GB)
Show the other 9 files
QuantizationFile sizeSmallest GPU
Q5_K17.8 GiBRTX 4090 (24 GB)
Q4_K15.4 GiBRTX 4090 (24 GB)
IQ3_S12.4 GiBRTX 4080 / 4080 SUPER (16 GB)
IQ3_XXS11.7 GiBRTX 4080 / 4080 SUPER (16 GB)
IQ2_S10.8 GiBRTX 4080 / 4080 SUPER (16 GB)
IQ2_XXS9.71 GiBRTX 4070 / 4070 SUPER (12 GB)
IQ2_M8.07 GiBRTX 4070 / 4070 SUPER (12 GB)
Q27.36 GiBRTX 4070 / 4070 SUPER (12 GB)
IQ1_M6.76 GiBRTX 4070 / 4070 SUPER (12 GB)

Read every day from Hugging Face. Split files are added together; vision projectors are left out.

How many GPUs to run Inkling 2 Mini

Fewest GPUs that fit the weights plus one 8K-token request. Data-center GPUs use vLLM (90% of memory usable), the others llama.cpp. Several GPUs split the model with tensor or pipeline parallelism.

GPUNVFP4 / MXFP8 experts + BF16 (native)BF16 / FP168-bit4-bit
GeForce RTX 4090 24 GBNot on this GPU321
GeForce RTX 5090 32 GBNot on this GPU211
RTX PRO 6000 Blackwell 96 GBNot on this GPU111
Mac M5 Max 128 GB 96 GB usableNot on this GPU111
Ryzen AI Max+ 395 128 GB 96 GB usableNot on this GPU111
DGX Spark (GB10) 128 GBNot on this GPU111
NVIDIA H100 80 GB1111
NVIDIA H200 141 GB1111
NVIDIA B200 (HGX) 180 GB1111

8-bit: FP8 with vLLM (INT8 on GPUs without FP8), Q8_0 with llama.cpp. 4-bit: AWQ / GPTQ with vLLM, Q4_K_M with llama.cpp, both counted at the Q4_K_M size. Not on this GPU: the official NVFP4 / MXFP8 experts + BF16 weights do not run there (llama.cpp needs a GGUF version, and vLLM needs a GPU with FP8 support).

Cheapest way to run Inkling 2 Mini today

For each GPU: the fewest cards that fit Inkling 2 Mini with one 8K-token request, the most faithful weight format at that count, and the cheapest on-demand price (2026-11-22).

On-demand prices from AWS, Google Cloud, Azure, RunPod, Vast.ai, Oracle Cloud, OVHcloud, Scaleway, Verda, Akash and UpCloud, checked every hour. Serving many users needs more KV cache, so more memory: size it in the calculator. Some links are affiliate links: StudioTV may earn a commission, at no extra cost to you.

AWS, Google Cloud and UpCloud prices: Spare Cores (opens in a new tab), CC BY-SA 4.0 (opens in a new tab), shared under the same license.

Or use Inkling 2 Mini through an API

No GPU to rent: these providers serve Inkling 2 Mini and bill per token. Speeds are measured by Hugging Face for one request.

ProviderOutput, per 1M tokensInput, per 1MSpeedFirst token
Primeintellect$0.300$0.07586 tokens/s1.1 s
Io-net$0.350$0.080111 tokens/s806 ms

Prices and speeds from Hugging Face Inference Providers (opens in a new tab), read every day (November 22, 2026). An API wins at low volume; renting GPUs wins when they stay busy: the calculator gives the break-even.

Recommended settings for Inkling 2 Mini

Sampling defaults from the official generation_config.json (opens in a new tab) of thinkingmachines/Inkling-2-Mini.

SettingValue
Temperature1

vLLM and SGLang apply these defaults unless a request sets its own. With llama.cpp, pass them yourself: --temp 1

Reasoning. Adjustable reasoning effort. Reasoning tokens take context and KV cache like any other token: leave room for them.

KV cache: how memory grows with context

Architecture: Looped MoE: 24 layers, the middle 12 run 3 times (48 layer visits, each with its own cache) · 128 experts · 8 global + 40 sliding-window visits (1024 tokens), GQA 8 KV × 128 · short convolutions · image input. Its KV cache is 2.6× smaller than if every layer used full attention, because most layers only keep a short window of recent tokens. Only 3.4B parameters are active per token, but all 26B must sit in memory: VRAM depends on the total, speed on the active part.

Context per requestBF16 cacheFP8 cache
8K tokens0.41 GiB0.20 GiB
32K tokens1.16 GiB0.58 GiB
128K tokens4.16 GiB2.08 GiB
1M tokens32.2 GiB16.1 GiB

Plus 3.00 MiB of recurrent state per request, whatever the context.

Inkling 2 Mini VRAM FAQ

How much VRAM does Inkling 2 Mini need?

In BF16 the weights alone take 49.2 GiB (53 GB). In FP8 that is 25.4 GiB, and about 15.8 GiB with 4-bit quantization (Q4_K_M); the official NVFP4 / MXFP8 experts + BF16 checkpoint is 18.2 GiB. Each request then adds KV cache: 0.41 GiB at 8K tokens and 1.16 GiB at 32K (BF16 cache).

Can Inkling 2 Mini run on a single RTX 4090 (24 GB)?

Yes. In Q4_K_M it fits on one RTX 4090 with an 8K-token context (16.3 GiB including the cache, with llama.cpp).

Can Inkling 2 Mini run on an RTX 5090 (32 GB)?

Yes. In Q8_0 (25.8 GiB with an 8K-token context) it fits on one RTX 5090, with llama.cpp.

Can Inkling 2 Mini run on a Mac?

Yes. In 4-bit (Q4_K_M with llama.cpp) it needs a Mac with 32 GB of unified memory or more, of which macOS gives the GPU about 21 GB, for one request with an 8K-token context. In 8-bit (Q8_0) it needs 48 GB or more.

Can Inkling 2 Mini run on a DGX Spark or a Ryzen AI Max+ 395 (128 GB)?

Yes. In BF16 (49.6 GiB with an 8K-token context) it fits on a DGX Spark (128 GB) and on a Ryzen AI Max+ 395 with 128 GB, whose GPU gets up to 96 GB, with llama.cpp and one request.

What is the cheapest way to run Inkling 2 Mini?

On 2026-11-22, the cheapest on-demand setup is 1 × GeForce RTX 3090 (24 GB) with Q4_K_M weights on Vast.ai, at $0.109 per hour (about $79.57 per month). Next: 2 × RTX A4000 (16 GB) with Q8_0 weights on Vast.ai, at $0.176 per hour. Sized for one 8K-token request; prices are checked every hour.

How many H100 GPUs does Inkling 2 Mini need?

With one request and an 8K-token context: 1 in BF16, 1 in FP8 and 1 in 4-bit (vLLM, 90% of memory usable). Serving many users at once needs more memory for their KV caches: the calculator sizes that for you.

Can Inkling 2 Mini read images?

Yes. Inkling 2 Mini reads text and images. Images add tokens to the context, so they also add KV cache.

Does Inkling 2 Mini support tool calling?

Yes, Inkling 2 Mini supports tool (function) calling: it can ask your application to run a function and use the result.

Does Inkling 2 Mini have a thinking mode?

Adjustable reasoning effort. Reasoning tokens count toward the context and the KV cache.

What temperature should I use with Inkling 2 Mini?

The official defaults are temperature 1.

Inkling 2 Mini VRAM badge

For a model card or a README: the 4-bit size and the smallest GPU it fits on, linked to this page. Data as JSON: /api/models/inkling-2-mini.json.

Inkling 2 Mini VRAM badge

[![Inkling 2 Mini VRAM](https://studiotvai.com/badge/inkling-2-mini.svg)](https://studiotvai.com/vram-requirements/inkling-2-mini)

Other models

Same family (sliding-window attention): Gemma 2 9B · Gemma 3 4B · Gemma 3 12B · Gemma 3 27B · Unlimited-OCR · Gemma 4 12B · Gemma 4 31B · Gemma 4 26B-A4B · gpt-oss 20B · gpt-oss 120B · Llama 4 Scout 17B-16E · Llama 4 Maverick 17B-128E · Kolibri-1 · Recursive Flint · Nemotron Labs Dory 1.0 4B · Step 3.5 Flash · Step 3.7 Flash · Laguna S 2.1 · Laguna XS 2.1 · Command A+ · Inkling Small · Muse Glimmer 30B · Humanizer · Mellum2.1 12B-A2.5B · Spark-X2.5 4B · Ministral 8B · Gemma 4 E4B · Gemma 4 E2B · Spark-X2.6 4B · Spark-X2.6 1.7B · Laguna XS.2 · Qwen3.8 27B-DSpine-b16 · Aion-1.0 2.3B

All 233 other models

Llama 3.1 8B · Llama 3 8B · Llama 3.2 3B · Nanbeige4.3 33B · Granite 4.2 30B · Granite 4.2 8B · Granite 4.2 3B · MiniCPM5 2B · MiniCPM5 1B · Ministral 3 14B · Ministral 3 8B · Ministral 3 3B · Devstral 2 123B · Devstral Small 2 24B · Mistral Medium 3.5 128B · Magistral Small 1.2 24B · Mistral Small 3 24B · Mistral Large 2411 123B · Pixtral Large 124B · Codestral 22B · Mistral Small 2409 22B · Pixtral 12B · Nemotron 3.5 Parse · Gemma 2 9B · Gemma 3 4B · Gemma 3 12B · Gemma 3 27B · LLaVA 1.6 Mistral 7B · Mistral Small 3.2 24B · Qwen3 32B · Llama 3.3 70B · Llama 3.1 70B · Llama 3.1 405B · Hy-Memory 30B-A3B · Kimi VL A3B Instruct · Mixtral 8x7B · Unlimited-OCR · Unlimited-OCR 2 · Xing4.0 29B-A4B · Qwen3.5 4B · MiniCPM-V-4.7 1B · Raon-1 mini preview · Raon-1 nano · Qwen3.5 9B · Qwen3.8 27B · Qwen3.6 35B-A3B · Maxwell V1 Flash · Clover 35B-A3B Preview · SenseNova Loop 9B Preview · Qwen3.5 122B-A10B · Qwen3.8-Flash-Next · Qwen3.8-Flash-Next E384 FP4 V2 · LFM3 1.7B · LFM3-VL 2B · d1 2B · LFM2.5 8B-A1B · LFM2 24B-A2B · Gemma 4 12B · Gemma 4 31B · Gemma 4 26B-A4B · gpt-oss 20B · gpt-oss 120B · Llama 4 Scout 17B-16E · Llama 4 Maverick 17B-128E · Kolibri-1 · Recursive Flint · Nemotron Labs Dory 1.0 4B · Step 3.5 Flash · Step 3.7 Flash · Laguna S 2.1 · Laguna XS 2.1 · Command A+ · Inkling Small · Muse Glimmer 30B · Nemotron 3 Nano 30B-A3B · Nemotron 3 Super 120B-A12B · Nemotron 3 Ultra 550B-A55B · Mamba-Codestral 7B · Nemotron 3.5 Super VL 120B-A12B · Nemotron 3.5 Bolt 74B-A9B · MiniMax M3.1 Flash · MiniMax M3 · MiniMax M2.7 · DeepSeek R1 · Kimi K2.6 · GLM-5.3 · Atria Dawn · Hydrogen-1.0 · Atria Dawn Preview · Hy4 Preview · Hy4 Flash · K-EXAONE 2.0 750B-A37B · Inkling · Inkling 2 Small · GLM-5.3-Flash · DeepSeek V4 Flash · DeepSeek V4 Pro · DeepSeek V4.1 Flash · LongCat-2.0 · LongCat-2.5 · ERNIE 5.5 Flash Base · MiMo-V2.6 Pro · MiMo-V2.6 Flash · Muse Spark Open · Beam 1.0 501B-A23B · Kimi K3 · Mistral Large 4 · Mistral Large 3 · Mistral Small 4 119B · Leanstral 1.5 119B-A6B · Mixtral 8x22B · Qwen3-Coder-Next 80B-A3B · Qwen3-Coder 30B-A3B · Qwen3-Coder 480B-A35B · DeepSeek V4 Flash Vision Exp · Qwen3.8 2.4T-A95B · Ornith 1.5 397B · Ornith 1.5 35B-A3B · Ornith 1.5 9B · Intern S2 397B · Clef · Clef Flash · d1 3B · Humanizer · JEV 27B VL · Mellum2.1 12B-A2.5B · LightOnOCR-3 4B · Spark-X2.5 4B · Agnes-3.0-Qwen · Youtu-Parsing-Omni · Agnes-3.0-Flash · Mistral 7B v0.3 · Mistral Small 3.1 24B · Mistral Nemo · Ministral 8B · DeepSeek V3 · DeepSeek R1 Qwen3 8B · DeepSeek R1 Distill Qwen 32B · DeepSeek V3.1 · DeepSeek R1 Distill Qwen 14B · DeepSeek R1 Distill Qwen 7B · DeepSeek R1 Distill Llama 8B · Qwen3 4B · Qwen3 8B · Qwen2.5 7B · Qwen3 14B · Qwen2.5 3B · Qwen3.6 27B · Qwen3.5 27B · Qwen2.5-Coder 7B · Qwen2.5 32B · Qwen2.5-Coder 14B · Qwen3.5 35B-A3B · Qwen3 30B-A3B · Qwen2.5 14B · Qwen2.5-Coder 32B · Kimi K2.7 Code · Kimi K2.5 · Kimi K2 · GLM-5.2 · GLM-5.1 · GLM-5 · MiniMax M2.5 · MiniMax M2 · MiMo-V2.5 · Ornith-1.0 397B · Qwen2.5-Coder 3B · Qwen3 235B-A22B · Qwen3.5 397B-A17B · Gemma 4 E4B · Gemma 4 E2B · Nemotron 3 Nano 4B · Nemotron 3.5 Lightning 30B-A3B · MiniCPM-V-4.7 35B-A3B · LFM2.5 24B-A2B · GLM-4.7 · Qwen3-VL 8B · Qwen2.5-VL 7B · Qwen3-VL 4B · Qwen2.5-VL 3B · Qwen2.5-VL 32B · Qwen2-VL 7B · GLM-4.5-Air · GLM-4.5 · phi-4 · Phi-3.5-mini · Phi-4-mini · Phi-3-mini-128k · Ling-3.1-flash · Step 5 Preview · Hy3 · Ornith-1.5 122B · Spark-X2.6 4B · Sage-A1.5 · MLExpert · Agents-A1 4B · Ling-3.0-flash-Sante · MiniCPM-OCR-1.0 · WeVisOne-VL 4B · Granite 4.1 3B · Qwen3-VL 30B-A3B · Mage-VL-Cowork · Spark-X2.6 1.7B · Qwen3.5 4B-Qwenputer · Kimi K3.1 Pro · 6.8-Code2Video · Qwen3.5 4B-GLB-3D-Coder · DeepSeek V3.2 · GLM-4.7-Flash · Laguna M.1 · Laguna XS.2 · Qwen3-VL 235B-A22B · Qwen2 7B · Qwen2.5 72B · Qwen3-VL 32B · Qwen2.5 14B-1M · Qwen3-Next 80B-A3B · Qwen2.5-VL 72B · Nemotron Nano 9B v2 · Qwen3.5 0.8B-weight-space-programming · Axion-1.0 35B · Hermes-3-Llama-3.1 8B · Qwen3.8 27B-DSpine-b16 · Granite 3.3 8B · MNN-Pocket 0.3B · S5-nano · Nex-N2.6 · Sage-S2.1 · KORMo-Web 11B · Hy3-preview · ERNIE 4.5 21B-A3B · Aion-1.0 2.3B · Hy4

Estimates, not guarantees: computed from the official config.json with the same engine as the LLM VRAM Calculator. Real usage depends on your engine version and settings.