Inkling Small — Hardware Requirements & GPU Compatibility
VisionChatFunctions
Inkling Small is a vision-language chat model from Thinking Machines, positioned as the smaller of two models in the Inkling family, though it still totals roughly 266 billion parameters. It accepts image input alongside text and is designed for general conversational use, combining visual understanding with standard dialogue capability. It is released under the Apache 2.0 license, permitting free local and commercial use. At this parameter scale, Inkling Small needs multi-GPU or server-class hardware to run at all; most users will reach it through a hosted endpoint rather than running it locally on their own systems.
Inkling Small runs on a 192 GB+ Mac or a datacenter GPU
Best local pick → Mac Pro M2 Ultra (192 GB) · Q4_K_M · 162.7 GB · ~98 tok/s
Macs from 96 GB (IQ2_XXS · ~108 tok/s)
Score 119.6#13 of 100 open#5 of 78 that fit ≤192 GBApache 2.0
Not set up for it?
Reference GPU · RTX 4090 24 GB
Estimates, ±20% · how we estimate
llmrun Composite Score →
119.6
- 90% range 117.2–121.9
- #13 of 100 open models
- #5 of 78 that fit ≤192 GB
Run Inkling Small
One recommended way. Every other method, OS and quant is in the step-by-step guide.
Recommended
ollama run hf.co/unsloth/Inkling-Small-GGUF:Q4_K_MQ4_K_M runs on one B200 192 GB · 16K context
Download: UD-Q4_K_M/Inkling-Small-UD-Q4_K_M-00001-of-00005.gguf162.54 GB + 4 partsunsloth
Other ways
each opens step-by-stepRun in cloud
Fits on B300 (288 GB) (125 GB headroom) · Q4_K_M
- Generation speed
- ~399 tok/s
- generation speed
- Cost per 1M output tokens
- $4.84
- per 1M output tokens
How Much VRAM Does Inkling Small Need?
Select a quantization level; the listed hardware adheres to it.
| Quantization | VRAM | +32K | Runs on | |
|---|---|---|---|---|
| 73.8 GB | 79.1 GB | 2-bit · i-quant extreme compression — Importance-weighted 2-bit, extreme compression — significant quality loss | Runs on H100 SXM 80 GB | |
| 90.4 GB | 95.7 GB | 2-bit · i-quant medium — Importance-weighted 2-bit, medium | Runs on RTX PRO 6000 Blackwell Max-Q Workstation Edition 96 GB | |
| 113.7 GB | 119 GB | 2-bit · K-quant — 2-bit quantization with K-quant improvements | Runs on Instinct MI250X 128 GB | |
| 130.3 GB | 135.6 GB | 3-bit medium — 3-bit medium quantization | Runs on H200 NVL 141 GB | |
| 160.2 GB | 165.5 GB | 4-bit medium — 4-bit medium quantization — most popular sweet spot | Runs on B200 192 GB | |
| 190.2 GB | 195.4 GB | 5-bit medium — 5-bit medium quantization — good quality/size tradeoff | Runs on Instinct MI325X 256 GB | |
| 220.1 GB | 225.4 GB | 6-bit, very good — 6-bit quantization, very good quality | Runs on Instinct MI325X 256 GB | |
| 266.6 GB | 271.9 GB | 8-bit, near-lossless — 8-bit quantization, near-lossless | Runs on Instinct MI350X 288 GB |
Selected Q4_K_M
4-bit medium quantization162.7GBat 16K context
- Weights
- 159.6 GB
- KV cache
- 2.8 GB
- Overhead
- 0.3 GB
Server-class: ~162.7 GB at 16K context · one B200 192 GB
Which Hardware Can Run Inkling Small?
Q4_K_M · 160.2 GBInkling Small (Q4_K_M) requires 160.2 GB of VRAM to load the model weights. For comfortable inference with headroom for KV cache and system overhead, 209+ GB is recommended. No consumer GPU has enough memory. 4 devices with unified memory can run Inkling Small, including Mac Studio (M3 Ultra, 256GB), Mac Studio (M3 Ultra, 512GB), Mac Pro M2 Ultra (192 GB).
128–144 GB
Q4_K_M · best fit IQ3_S (116.2 GB)
4 GPUs · 2 Mac classes · 1 device family
GH200 Grace Hopper Superchip~394Simulate output on GH200 Grace Hopper SuperchipH200 NVL~393Simulate output on H200 NVLH200 SXM~393Simulate output on H200 SXMM2 Ultra*128 GB~103Simulate output on Mac Pro M2 Ultra (128 GB)M5 Max128 GB~99Simulate output on Mac Studio M5 Max (128 GB)GB10~155Simulate output on NVIDIA DGX Spark192+ GB
fits · 29.3 GB free · up to 32K ctx
6 GPUs · 1 Mac class · 2 device families
B200~399Simulate output on B200B300~399Simulate output on B300Instinct MI350X~116Simulate output on Instinct MI350XM3 Ultra*256–512 GB~98Simulate output on Mac Studio (M3 Ultra, 256GB)8x H100 SXM5~422Simulate output on NVIDIA DGX H1008x A100 SXM4~415Simulate output on NVIDIA DGX A100 640GB
By VRAM tier · typical GPUs and Macs at Q4_K_M
Consumer GPUs (up to 32 GB) and Macs up to 64 GB: too small for this model (needs ≥ 76.3 GB)8 GB tier →12 GB tier →16 GB tier →24 GB tier →Macs →
How good is Inkling Small?
The llmrun Score merges 11 public benchmark boards into a single figure, where 100 represents the average for open models.
Score vs VRAM in the ≤192 GB tier
- Inkling Small
- Other models
- Best Score for its VRAM
#5 of 78 that fit ≤192 GB
Each dot is a model at the best quant that fits ≤192 GB. The line joins the models with the best Score for their VRAM; a dot below it is beaten by a model that needs no more memory.
Full ≤192 GB ranking →Benchmarks
The 6 boards that weigh most in the llmrun Score, of 13 with a result for this model; badges show who ran the test.
- GPQA Diamond
Epoch AI · Jul 2026Independent
Independent88.5%
#13/98 open
- LMArena Text
LMArena · Oct 2026Arena
Arena1404.0
#46/187 open
- AIME 2024/2025
Epoch AI · Jul 2026Independent
Independent90.0%
#14/86 open
- FrontierMath
Epoch AI · Jul 2026Independent
Independent46.3%
#11/17 open
- LMArena WebDev
LMArena · Oct 2026Arena
Arena1409.1
#27/52 open
- ARC-AGI-2
Epoch AI · Jul 2026Independent
Independent40.1%
#7/19 open
In the ≤192 GB class
78 models fit ≤192 GBSame Score, only models that fit ≤192 GB at their best quant: the ranking a ≤192 GB owner actually chooses from.
- #1
MiMo V2.6 Flash RLXiaomi
131.9
~229 tok/s · 186.5 GB
- #2
DeepSeek V4 Flash 0731DeepSeek
127.6
~181 tok/s · 183.0 GB
- #3
GLM 5.3 FlashZ.ai
124.2
~228 tok/s · 185.1 GB
- #4
Qwen3.8 27BAlibaba
122.1
~299 tok/s · 17.9 GB
- #6
Qwen3.5 397B A17BAlibaba
113.0
~185 tok/s · 182.1 GB
Alternatives to Inkling Small
Inkling family · 3 models →Official models with an llmrun Score, each answering one question; sizes are each model's default quant (usually Q4_K_M) with a 16K-token context.
Best alternative at this size
MiMo V2.6 Flash RL
Xiaomi
- 12.3 higher Score
- 23.8 GB more VRAM
Score 131.9 · 186.5 GB
Other vendor
DeepSeek V4 Flash 0731
DeepSeek
- 8 higher Score
- 20.3 GB more VRAM
Score 127.6 · 183 GB
Similar models
10| Model | Score | VRAM | tok/s | Why it is similar |
|---|---|---|---|---|
| GLM 5.3 FlashZ.ai | 124.2 | 205.2 GB | —needs > 24 GB | similar score |
| Qwen3.8 27BAlibaba | 122.1 | 17.9 GB | ~38 | similar score |
| Qwen3.5 397B A17BAlibaba | 113.0 | 242.6 GB | —needs > 24 GB | similar score |
| MiniMax M3MiniMax | 112.5 | 258 GB | —needs > 24 GB | similar score |
| MiniMax M2.5MiniMax | 111.8 | 139.6 GB | —needs > 24 GB | similar score |
| Qwen3.6 27BAlibaba | 111.2 | 17.9 GB | ~38 | similar score |
| DeepSeek V4 FlashDeepSeek | 109.3 | 175.1 GB | —needs > 24 GB | similar score |
| GLM 4.7Z.ai | 109.1 | 217.9 GB | —needs > 24 GB | similar score |
| Gemma 4 31B ITGoogle | 106.8 | 21.7 GB | ~32 | similar score |
| MiMo V2.5Xiaomi | 108.3 | 187.6 GB | —needs > 24 GB | same size |
Specifications
- Publisher
- Thinking Machines
- Family
- Inkling
- Parameters
- 266.0B
- Active parameters
- 1.7B per token
- Architecture
- InklingForConditionalGeneration
- Vocabulary Size
- 201,024
- Release Date
- 2026-07-27
- License
- Apache 2.0
Frequently Asked Questions
How much VRAM does Inkling Small need?
Inkling Small requires 160.2 GB of VRAM at Q4_K_M, or 532.6 GB at BF16.
VRAM = Weights + KV Cache + Overhead
Weights = 266.0B × 4.8 bits ÷ 8 = 159.6 GB
KV Cache + Overhead ≈ 0.6 GB (at 2K context + ~0.3 GB framework)
Fit ratings and hardware model lists check this model with room for a 16K-token context, which needs a little more memory.
VRAM usage by quantization
Can NVIDIA GeForce RTX 5090 run Inkling Small?
No — Inkling Small requires at least 73.8 GB at IQ2_XXS, which exceeds the NVIDIA GeForce RTX 5090's 32 GB of VRAM.
What's the best quantization for Inkling Small?
For Inkling Small, Q4_K_M (160.2 GB) offers the best balance of quality and VRAM usage. Q5_K_S (183.5 GB) provides better quality if you have the VRAM. The smallest option is IQ2_XXS at 73.8 GB.
VRAM requirement by quantization
★ Recommended — best balance of quality and VRAM usage.
Can I run Inkling Small on a Mac?
Yes — Mac Studio (M3 Ultra, 96GB) and 20 other Macs can run Inkling Small, the smallest at IQ2_XXS. Apple Silicon uses unified memory, so the model shares RAM with the system. At Q4_K_M you need at least 160.2 GB of usable unified memory (RAM minus macOS overhead), from Mac Pro M2 Ultra (192 GB).
Can I run Inkling Small locally?
Yes — Inkling Small can run locally on consumer hardware. Popular tools include Ollama, LM Studio, and llama.cpp.
How fast is Inkling Small?
At Q4_K_M, Inkling Small can reach ~116 tok/s on AMD Instinct MI350X. Speed depends mainly on GPU memory bandwidth. Real-world results typically within ±20%.
tok/s = 1000 ÷ (active GB ÷ (bandwidth GB/s × efficiency) × 1000 + layers × routing ms)
Mixture-of-Experts: just the active experts get read for each token, along with a fixed routing cost per layer.
Example: NVIDIA B200 → 1.0 GB active ÷ (8000 × 0.65) = 0.199 ms, plus 42 layers × 0.055 ms = 2.310 ms, so 1000 ÷ 2.509 ms = ~399 tok/s
Estimated speed at Q4_K_M (160.2 GB)
Real-world results typically within ±20%. Speed depends on batch size, quantization kernel, and software stack.
What's the download size of Inkling Small?
At Q4_K_M, the download is about 159.57 GB. The full-precision BF16 version is 531.91 GB. The smallest option (IQ2_XXS) is 73.14 GB.
Which GPUs can run Inkling Small?
No single consumer GPU has enough VRAM to run Inkling Small at Q4_K_M (160.2 GB). Multi-GPU or professional hardware is required.
Which devices can run Inkling Small?
4 devices with unified memory can run Inkling Small at Q4_K_M (160.2 GB), including Mac Pro M2 Ultra (192 GB), Mac Studio (M3 Ultra, 256GB), Mac Studio (M3 Ultra, 512GB), Mac Studio M2 Ultra (192 GB). Apple Silicon Macs use unified memory shared between CPU and GPU, making them well-suited for local LLM inference.