Cosmos Reason2 2B — Hardware Requirements & GPU Compatibility
VisionChatReasoningFunctions
Cosmos Reason2 2B is NVIDIA's 2.4-billion-parameter open reasoning vision-language model for physical AI, built on the Qwen3-VL-2B-Instruct backbone and designed to let robots and embodied agents perceive a scene and reason step by step about space, time, and physical common sense before acting. It generates long chains of thought and can output structured spatial data such as 2D/3D point localization, bounding boxes, and trajectories, plus OCR reading of scene text, going beyond a general chat assistant into physical-world planning. At just 2.4 billion parameters it is small enough to run on a single consumer GPU. Context length is 256,000 tokens, up sharply from the 16,000 tokens of the original Cosmos Reason1. It is released under the NVIDIA Open Model License, a custom license permitting commercial use and redistribution provided derivatives carry a "Built on NVIDIA Cosmos" attribution notice. It was published in December 2025, alongside a larger 8B sibling.
Runs great on a 24 GB GPU
Runs from 8 GB (Q4_K_M · 32K)Best on 24 GB (Q4_K_M · ~407 tok/s · 32K)Macs from 8 GB (Q4_K_M · ~45 tok/s)
Other
Reference GPU · RTX 4090 24 GB
Estimates, ±20% · how we estimate
Run Cosmos Reason2 2B
One recommended way. Every other method, OS and quant is in the step-by-step guide.Each way opens a step-by-step guide for your OS and quant.
Recommended
ollama run hf.co/apolo13x/Cosmos-Reason2-2B-GGUF:Q4_K_MQ4_K_M fits fully on the RTX 4090 · 20.5 GB free · 16K context
Download: Cosmos-Reason2-2B-Q4_K_M.gguf1.28 GBapolo13x
Ways to runOther ways
each opens step-by-stepRun in cloud
Runs on RTX 3060 12GB (8 GB headroom) · Q4_K_M
- Generation speed
- ~145 tok/s
- generation speed
- Cost per 1M output tokens
- $0.11
- per 1M output tokens
How Much VRAM Does Cosmos Reason2 2B Need?
Select a quantization level; the listed hardware adheres to it.
| Quantization | VRAM | +32K | On RTX 4090 (reference) | |
|---|---|---|---|---|
| 1.1 GB | 5.1 GB | 2-bit · K-quant — 2-bit quantization with K-quant improvements | Fits: 32K context · ~575 tok/s | |
| 1.3 GB | 5.3 GB | 3-bit medium — 3-bit medium quantization | Fits: 32K context · ~500 tok/s | |
| 1.6 GB | 5.6 GB | 4-bit medium — 4-bit medium quantization — most popular sweet spot | Fits: 32K context · ~407 tok/s | |
| 1.9 GB | 5.9 GB | 5-bit medium — 5-bit medium quantization — good quality/size tradeoff | Fits: 32K context · ~343 tok/s | |
| 2.2 GB | 6.2 GB | 6-bit, very good — 6-bit quantization, very good quality | Fits: 32K context · ~297 tok/s | |
| 2.7 GB | 6.6 GB | 8-bit, near-lossless — 8-bit quantization, near-lossless | Fits: 32K context · ~245 tok/s | |
| 5.4 GB | 9.3 GB | 16-bit · bfloat — Brain floating point 16 — preferred for training | Fits: 32K context · ~122 tok/s |
Selected Q4_K_M
4-bit medium quantization3.5GBat 16K context
- Weights
- 1.1 GB
- KV cache
- 2.1 GB
- Overhead
- 0.3 GB
Fits RTX 4090 (reference) · 20.6 GB free · ~407 tok/s
≥50 tok/s needs ≥124 GB/s · RTX 5090 ~724, RTX 3090 Ti ~407
Which Hardware Can Run Cosmos Reason2 2B?
Q4_K_M · 1.6 GBCosmos Reason2 2B (Q4_K_M) requires 1.6 GB of VRAM to load the model weights. For comfortable inference with headroom for KV cache and system overhead, 3+ GB is recommended. 52 GPUs can run it, including NVIDIA GeForce RTX 5090, NVIDIA GeForce RTX 3090 Ti. 163 devices with unified memory can run Cosmos Reason2 2B, including Mac Studio (M3 Ultra, 256GB), Mac Studio (M3 Ultra, 512GB).
≤ 11 GB
≤ 11 GB tier →fits · 4.5 GB free · up to 32K ctx
16 GPUs · 1 Mac class · 2 device families
RTX 3080~307Simulate output on RTX 3080RTX 2080 Ti~249Simulate output on RTX 2080 TiRTX 3070 Ti~246Simulate output on RTX 3070 TiM3*8 GB~45Simulate output on iMac 24" M3 (8 GB)Orin Nano~41Simulate output on NVIDIA Jetson Orin Nano 8GB (Super)iPhone~30Simulate output on iPhone 1720 GB
fits · 16.6 GB free · up to 32K ctx
1 GPU
RX 7900 XT~298Simulate output on RX 7900 XT32–40 GB
fits · 28.6 GB free · up to 32K ctx
1 GPU · 3 Mac classes · 4 device families
RTX 5090~724Simulate output on RTX 5090M5 Max36 GB~200Simulate output on Mac Studio M5 Max (36 GB)M532 GB~67Simulate output on MacBook Air 13" M5 (32 GB)Orin~83Simulate output on NVIDIA Jetson AGX Orin 32GBSnapdragon X Elite (X1E-80-100)~55Simulate output on Snapdragon X Elite Copilot+ PC
By VRAM tier · typical GPUs and Macs at Q4_K_M
Similar models
6| Model | Score | VRAM | tok/s | Why it is similar |
|---|---|---|---|---|
| MiniCPM5 2BOpenBMB | — | 2.5 GB | ~345 | same size |
| North Micro Vision InstructCohere | — | 3.7 GB | ~323 | same size |
| Qwen3.5 2BAlibaba | — | 1.9 GB | ~370 | same size |
| Granite 3.1 2B InstructIBM | — | 3.2 GB | ~329 | same size |
| Gemma 2 2B ITGoogle | — | 2.5 GB | ~379 | same size |
| EXAONE 3.5 2.4B InstructLGAI-EXAONE | — | 3.4 GB | ~412 | same size |
Frequently Asked Questions
How much VRAM does Cosmos Reason2 2B need?
Cosmos Reason2 2B requires 1.6 GB of VRAM at Q4_K_M, or 5.4 GB at BF16.
VRAM = Weights + KV Cache + Overhead
Weights = 2.4B × 4.8 bits ÷ 8 = 1.5 GB
KV Cache + Overhead ≈ 0.1 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
What's the best quantization for Cosmos Reason2 2B?
For Cosmos Reason2 2B, Q4_K_M (1.6 GB) offers the best balance of quality and VRAM usage. Q5_K_M (1.9 GB) provides better quality if you have the VRAM. The smallest option is Q2_K at 1.1 GB.
VRAM requirement by quantization
★ Recommended — best balance of quality and VRAM usage.
Can I run Cosmos Reason2 2B on a Mac?
Yes — MacBook Air 13" M3 (8 GB) and 145 other Macs can run Cosmos Reason2 2B. Apple Silicon uses unified memory, so the model shares RAM with the system. At Q4_K_M you need at least 1.6 GB of usable unified memory (RAM minus macOS overhead).
Can I run Cosmos Reason2 2B locally?
Yes — Cosmos Reason2 2B can run locally on consumer hardware. Popular tools include Ollama, LM Studio, and llama.cpp.
How fast is Cosmos Reason2 2B?
At Q4_K_M, Cosmos Reason2 2B can reach ~2981 tok/s on AMD Instinct MI350X. On NVIDIA GeForce RTX 4090: ~407 tok/s. Speed depends mainly on GPU memory bandwidth. Real-world results typically within ±20%.
tok/s = (bandwidth GB/s ÷ model GB) × efficiency
Example: NVIDIA B200 → 8000 ÷ 1.61 × 0.65 = ~3230 tok/s
Estimated speed at Q4_K_M (1.6 GB)
Real-world results typically within ±20%. Speed depends on batch size, quantization kernel, and software stack.
What's the download size of Cosmos Reason2 2B?
At Q4_K_M, the download is about 1.46 GB. The full-precision BF16 version is 4.88 GB. The smallest option (Q2_K) is 1.04 GB.
Which GPUs can run Cosmos Reason2 2B?
52 consumer GPUs can run Cosmos Reason2 2B at Q4_K_M. Top options include AMD Radeon RX 6700 XT, AMD Radeon RX 6800, AMD Radeon RX 6800 XT. 52 GPUs have plenty of headroom for comfortable inference.
Which devices can run Cosmos Reason2 2B?
163 devices with unified memory can run Cosmos Reason2 2B at Q4_K_M (1.6 GB), including AMD Ryzen AI 9 HX 370 (Strix Point) Laptop, ASUS Ascent GX10, Beelink GTR9 Pro (Ryzen AI Max+ 395, 128 GB), Framework Desktop (Ryzen AI Max+ 395, 128 GB). Apple Silicon Macs use unified memory shared between CPU and GPU, making them well-suited for local LLM inference.