Cosmos Reason2 8B — Hardware Requirements & GPU Compatibility
VisionChatReasoningFunctions
Cosmos Reason2-8B is NVIDIA's 8.8-billion-parameter open reasoning vision-language model for physical AI, built on a Qwen3-VL-8B-Instruct backbone and tuned to reason step by step about video and images the way a human would when planning actions in the real world. Rather than just labeling objects, it applies physics, spatio-temporal understanding, and common sense to tasks like robot planning, autonomous-vehicle video captioning, and video-analytics annotation, producing structured outputs such as 2D/3D point localization, bounding boxes, trajectory coordinates, and on-screen OCR text. It ships alongside a smaller 2B variant for edge deployment, while the 8B model needs a capable single GPU or more, less once quantized. Context length is roughly 256,000 tokens, up sharply from 16,000 tokens in the original Cosmos Reason 1. It is released under the NVIDIA Open Model License, a custom license that permits commercial use and derivative models but requires attribution ("Built on NVIDIA Cosmos") and prohibits removing its safety guardrails. It was published in December 2025.
Runs great on a 24 GB GPU
Runs from 8 GB (Q4_K_M · 8K)Best on 24 GB (Q4_K_M · ~113 tok/s · 32K)Macs from 16 GB (Q4_K_M · ~24 tok/s)
Other
Reference GPU · RTX 4090 24 GB
Estimates, ±20% · how we estimate
Run Cosmos Reason2 8B
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/mradermacher/Cosmos-Reason2-8B-GGUF:Q4_K_MQ4_K_M fits fully on the RTX 4090 · 15.6 GB free · 16K context
Download: Cosmos-Reason2-8B.Q4_K_M.gguf5.03 GBmradermacher
Ways to runOther ways
each opens step-by-stepRun in cloud
Runs on RTX 3080 Ti (12 GB) (3 GB headroom) · Q4_K_M
- Generation speed
- ~102 tok/s
- generation speed
- Cost per 1M output tokens
- $0.33
- per 1M output tokens
How Much VRAM Does Cosmos Reason2 8B Need?
Select a quantization level; the listed hardware adheres to it.
| Quantization | VRAM | +32K | On RTX 4090 (reference) | |
|---|---|---|---|---|
| 4.1 GB | 9.7 GB | 2-bit · K-quant — 2-bit quantization with K-quant improvements | Fits: 32K context · ~160 tok/s | |
| 4.2 GB | 9.9 GB | 3-bit small — 3-bit small quantization | Fits: 32K context · ~155 tok/s | |
| 4.7 GB | 10.3 GB | 3-bit medium — 3-bit medium quantization | Fits: 32K context · ~139 tok/s | |
| 4.8 GB | 10.5 GB | 4-bit · legacy — 4-bit legacy quantization | Fits: 32K context · ~136 tok/s | |
| 5.8 GB | 11.4 GB | 4-bit medium — 4-bit medium quantization — most popular sweet spot | Fits: 32K context · ~113 tok/s | |
| 6.9 GB | 12.5 GB | 5-bit medium — 5-bit medium quantization — good quality/size tradeoff | Fits: 32K context · ~95 tok/s | |
| 8.0 GB | 13.6 GB | 6-bit, very good — 6-bit quantization, very good quality | Fits: 32K context · ~82 tok/s | |
| 9.6 GB | 15.3 GB | 8-bit, near-lossless — 8-bit quantization, near-lossless | Fits: 32K context · ~68 tok/s |
Selected Q4_K_M
4-bit medium quantization8.4GBat 16K context
- Weights
- 5.1 GB
- KV cache
- 3.0 GB
- Overhead
- 0.3 GB
Fits RTX 4090 (reference) · 15.6 GB free · ~113 tok/s
≥50 tok/s needs ≥445 GB/s · RTX 5090 ~201, RTX 3090 Ti ~113
Which Hardware Can Run Cosmos Reason2 8B?
Q4_K_M · 5.8 GBCosmos Reason2 8B (Q4_K_M) requires 5.8 GB of VRAM to load the model weights. For comfortable inference with headroom for KV cache and system overhead, 8+ GB is recommended. 52 GPUs can run it, including NVIDIA GeForce RTX 5090, NVIDIA GeForce RTX 3090 Ti, NVIDIA GeForce RTX 3070 Ti. 151 devices with unified memory can run Cosmos Reason2 8B, including Mac Studio (M3 Ultra, 256GB), Mac Studio (M3 Ultra, 512GB), NVIDIA Jetson Orin Nano 8GB (Super).
12 GB
12 GB tier →fits · 3.6 GB free · up to 32K ctx
10 GPUs · 1 device family
RTX 3080 Ti~102Simulate output on RTX 3080 TiRTX 5070~75Simulate output on RTX 5070RTX 4070~57Simulate output on RTX 4070iPhone~9Simulate output on Apple iPhone 17 Pro20 GB
fits · 11.6 GB free · up to 32K ctx
1 GPU
RX 7900 XT~83Simulate output on RX 7900 XT32–40 GB
fits · 23.6 GB free · up to 32K ctx
1 GPU · 3 Mac classes · 4 device families
RTX 5090~201Simulate output on RTX 5090M5 Max36 GB~56Simulate output on Mac Studio M5 Max (36 GB)M532 GB~19Simulate output on MacBook Air 13" M5 (32 GB)Orin~23Simulate output on NVIDIA Jetson AGX Orin 32GBSnapdragon X Elite (X1E-80-100)~15Simulate 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 |
|---|---|---|---|---|
| Granite Switch 4.1 8BIBM | — | 8.4 GB | ~111 | same size |
| Apertus V1.5 8Bswiss-ai | — | 8.5 GB | ~111 | same size |
| Ornith 1.0 9BDeep Reinforce | — | 6.5 GB | ~106 | same size |
| LFM2.5 8B A1BLiquid AI | — | 5.6 GB | ~335 | same size |
| MiniCPM O 4 5OpenBMB | — | 8.3 GB | ~105 | same size |
| Qwen3 VL 8B InstructAlibaba | — | 8 GB | ~112 | same size |
Frequently Asked Questions
How much VRAM does Cosmos Reason2 8B need?
Cosmos Reason2 8B requires 5.8 GB of VRAM at Q4_K_M, or 19.3 GB at BF16.
VRAM = Weights + KV Cache + Overhead
Weights = 8.8B × 4.8 bits ÷ 8 = 5.3 GB
KV Cache + Overhead ≈ 0.5 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 8B?
For Cosmos Reason2 8B, Q4_K_M (5.8 GB) offers the best balance of quality and VRAM usage. Q5_K_S (6.6 GB) provides better quality if you have the VRAM. The smallest option is IQ2_XXS at 2.6 GB.
VRAM requirement by quantization
★ Recommended — best balance of quality and VRAM usage.
Can I run Cosmos Reason2 8B on a Mac?
Yes — Mac Mini M2 Pro (16 GB) and 133 other Macs can run Cosmos Reason2 8B. Apple Silicon uses unified memory, so the model shares RAM with the system. At Q4_K_M you need at least 5.8 GB of usable unified memory (RAM minus macOS overhead).
Can I run Cosmos Reason2 8B locally?
Yes — Cosmos Reason2 8B can run locally on consumer hardware. Popular tools include Ollama, LM Studio, and llama.cpp.
How fast is Cosmos Reason2 8B?
At Q4_K_M, Cosmos Reason2 8B can reach ~829 tok/s on AMD Instinct MI350X. On NVIDIA GeForce RTX 4090: ~113 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 ÷ 5.79 × 0.65 = ~898 tok/s
Estimated speed at Q4_K_M (5.8 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 8B?
At Q4_K_M, the download is about 5.26 GB. The full-precision BF16 version is 17.53 GB. The smallest option (IQ2_XXS) is 2.41 GB.
Which GPUs can run Cosmos Reason2 8B?
52 consumer GPUs can run Cosmos Reason2 8B at Q4_K_M. Top options include AMD Radeon RX 6700 XT, AMD Radeon RX 6800, AMD Radeon RX 6800 XT, AMD Radeon RX 7600. 40 GPUs have plenty of headroom for comfortable inference.
Which devices can run Cosmos Reason2 8B?
151 devices with unified memory can run Cosmos Reason2 8B at Q4_K_M (5.8 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.