text model · Flint · Windows
Can I run Flint 40B-A2.5B on AMD Radeon RX 9070 XT (16GB)?
No. Flint 40B-A2.5B needs ~25.9 GB even at Q4_K_M, but AMD Radeon RX 9070 XT (16GB) only has ~15 GB usable.
Needs ~25.9 GB even at Q4_K_M, but only ~15 GB is usable.
That figure is at a 4k context and moves about ±15% as context length changes.
The gap is about 10.9 GB: Flint 40B-A2.5B needs roughly 25.9 GB at Q4_K_M and AMD Radeon RX 9070 XT (16GB) leaves only about 15 GB usable for a model. The lightest tracked hardware that runs Flint 40B-A2.5B is the Nvidia GeForce RTX 5090 (32GB) at 32 GB. See Flint 40B-A2.5B on Nvidia GeForce RTX 5090 (32GB).
- Q4_K_M needed
- ~25.9 GB
- Usable on device
- ~15 GB
- Device memory
- 16 GB
Which quant fits
- Parameters
- 39.8B (MoE, 2.5B active)
- Q4_K_M size
- 23.61 GB
- Q8_0 size
- 39.44 GB
- Context
- 256k
- Ollama tag
- flint:40b
- Memory
- 16 GB vram
- Usable for weights
- ~15 GB
- Power draw
- ~304 W
- Best runtime
- llama.cpp (Vulkan/ROCm) / Ollama (ROCm)
What you can run instead
Run Flint 40B-A2.5B on other hardware
FAQ
Can AMD Radeon RX 9070 XT (16GB) run Flint 40B-A2.5B?
No. Flint 40B-A2.5B needs ~25.9 GB even at Q4_K_M, but AMD Radeon RX 9070 XT (16GB) only has ~15 GB usable.
How much memory does Flint 40B-A2.5B need?
AMD Radeon RX 9070 XT (16GB) does not have enough memory. At Q4_K_M the weights are ~23.61 GB; with KV cache and runtime overhead, budget ~25.9 GB at a 4k context. It is a Mixture-of-Experts model (39.8B total / 2.5B active), so all experts must stay in memory; memory tracks total params, not active params.
What is the best tool to run Flint 40B-A2.5B on Windows?
LM Studio for a simple setup; Ollama (CUDA) for the most speed. AMD GPUs run via Vulkan/ROCm at roughly half CUDA throughput. NVIDIA is the smooth path on Windows.
Embed this
A live badge for your model card or README, updated as the data is. [](https://localmodel.run/can-i-run/flint-40b-a2.5b/amd-rx-9070-xt-16gb) Sources
-
amd.com · 1 source
-
en.wikipedia.org · 1 source
-
hothardware.com · 1 source
-
huggingface.co · 3 sources
-
lmstudio.ai · 1 source
-
ollama.com · 2 sources
Weights are measured from GGUF files; KV cache and overhead are computed, so totals can vary ~15% with context and runtime. Any tok/s is a bandwidth estimate. See methodology.