Skip to content
RigCheckWhat can I run?

What can a Radeon VII / Instinct MI50 16 GB run locally?

What runs on a Radeon VII / Instinct MI50 16 GB? Every answer below is the one RigCheck's own checker for that model gives for this card, with 32 GB or 64 GB of system RAM. Nothing on this page is judged separately.

With 64 GB of system RAM, the checkers have an answer for a Radeon VII / Instinct MI50 16 GB (16 GB of VRAM) for 7 of the 9 models, sizes and routes they cover: Qwen3.8-Flash-Next in the Strata engine, Qwen3.8-27B in llama.cpp, Gemma 4 31B, Gemma 4 26B-A4B, Gemma 4 12B, Gemma 4 E4B, Gemma 4 E2B with caveats (part in system RAM, an experimental path, a planning budget or an unmeasured peak). With 32 GB, 0 without caveats. The llama.cpp language-model answers are capacity estimates and the Nemotron 3.5, GLM-5.3-Flash and DeepSeek V4.1 Flash answers are published route matches, not measured setups.

Radeon VII / Instinct MI50 16 GB: model by model

Language models

Model (checker)With 32 GB RAMWith 64 GB RAM
Qwen3.8-Flash-Next in the Strata engine! Experimental path
IQ3_XXS
Linux only
No measured speed for this hardware and setup yet.
Your RAM cannot hold this size's experts (about 53 GB of RAM), so Strata switches to low-RAM mode by itself and reads the experts the card does not hold from the SSD while it answers. Strata: slower, and much slower with a small graphics card; RigCheck has no matching speed for this. For code, the Coder (IQ1_M) runs here with its experts copied into RAM at start; Strata: it is weaker outside code and in languages other than English. Covered by the community "Older GPUs" guide (Linux). The published MI50 benchmark used two cards. Its gfx906 build failed from engine 0.1.44 to 0.2.0; 0.2.1 builds again but has not been run on a card (AMD_HIP.md).		! Experimental path
IQ3_S
Linux only
No measured speed for this hardware and setup yet.
Fits by the docs' rule (experts + ~10 GB), with little room to spare: close the browser while it runs. Covered by the community "Older GPUs" guide (Linux). The published MI50 benchmark used two cards. Its gfx906 build failed from engine 0.1.44 to 0.2.0; 0.2.1 builds again but has not been run on a card (AMD_HIP.md).
Qwen3.8-27B in llama.cpp! Fits with caveats (capacity estimate)
UD-Q4_K_XL · system RAM, CPU only
No measured speed for this hardware and setup yet.
RigCheck capacity estimate, not a measured setup.		! Fits with caveats (capacity estimate)
Q8_0 · system RAM, CPU only
No measured speed for this hardware and setup yet.
RigCheck capacity estimate, not a measured setup.
Gemma 4 (E2B, E4B, 12B, 26B-A4B, 31B) in llama.cpp! Best size: fits with caveats (capacity estimate)
Gemma 4 31B: fits with caveats (capacity estimate) (IQ4_XS, system RAM, CPU only); Gemma 4 26B-A4B: fits with caveats (capacity estimate) (UD-Q4_K_XL, system RAM, CPU only); Gemma 4 12B: fits with caveats (capacity estimate) (Q8_0, system RAM, CPU only); Gemma 4 E4B: fits with caveats (capacity estimate) (BF16, system RAM, CPU only); Gemma 4 E2B: fits with caveats (capacity estimate) (BF16, system RAM, CPU only)
No measured speed for this hardware and setup yet.
RigCheck capacity estimate, not a measured setup.		! Best size: fits with caveats (capacity estimate)
Gemma 4 31B: fits with caveats (capacity estimate) (Q8_0, system RAM, CPU only); Gemma 4 26B-A4B: fits with caveats (capacity estimate) (BF16, system RAM, CPU only); Gemma 4 12B: fits with caveats (capacity estimate) (BF16, system RAM, CPU only); Gemma 4 E4B: fits with caveats (capacity estimate) (BF16, system RAM, CPU only); Gemma 4 E2B: fits with caveats (capacity estimate) (BF16, system RAM, CPU only)
No measured speed for this hardware and setup yet.
RigCheck capacity estimate, not a measured setup.
No answer here: Qwen3.8-Flash-Next in llama.cpp, Mistral Medium 3.5 in llama.cpp, Mistral Large 4, Kolibri-1 (official FP8 route), Nemotron 3.5 (Bolt 74B-A9B, Super VL 120B-A12B), GLM-5.3-Flash (Unsloth GGUF), DeepSeek V4.1 Flash (September release, 1017 update) in DwarfStar
Model (checker)With 32 GB RAMWith 64 GB RAM
Qwen3.8-Flash-Next in llama.cpp✕ Does not fit
Even the smallest file (UD-IQ1_S, 72.5 GB) needs about 76.1 GB in system RAM. This checker does not cover your graphics card, so it is not counted; its memory could change the answer.		✕ Does not fit
Even the smallest file (UD-IQ1_S, 72.5 GB) needs about 76.1 GB in system RAM. This checker does not cover your graphics card, so it is not counted; its memory could change the answer.
Mistral Medium 3.5 in llama.cpp– Not covered– Not covered
Mistral Large 4… Weights not released
smallest estimate ~210 GB vs 48 GB of memory (VRAM + RAM, or a Mac's unified memory)		… Weights not released
smallest estimate ~210 GB vs 80 GB of memory (VRAM + RAM, or a Mac's unified memory)
Kolibri-1 (official FP8 route)– Not covered– Not covered
Nemotron 3.5 (Bolt 74B-A9B, Super VL 120B-A12B)– Not covered– Not covered
GLM-5.3-Flash (Unsloth GGUF)– Not covered– Not covered
DeepSeek V4.1 Flash (September release, 1017 update) in DwarfStar– Not covered– Not covered

Image models

No answer in this setup: Qwen-Image-2.1 (image generation), FLUX 3 Image Klein (int8, ComfyUI)
Model (checker)With 32 GB RAMWith 64 GB RAM
Qwen-Image-2.1 (image generation)– Not covered– Not covered
FLUX 3 Image Klein (int8, ComfyUI)– Not covered– Not covered

Video models

No answer in this setup: MiniMax H3 (video + audio)
Model (checker)With 32 GB RAMWith 64 GB RAM
MiniMax H3 (video + audio)– Not covered– Not covered

Quick setup, as in the home page finder: one Radeon VII / Instinct MI50 16 GB, a 32K context and no image input for the llama.cpp capacity estimates (Nemotron 3.5, GLM-5.3-Flash and DeepSeek V4.1 Flash are matched against published recipes, memory tables and routes at their own context instead (NVIDIA's 64K / 256K recipes; Unsloth's GLM table states no context length; DwarfStar's routes per DeepSeek checkpoint (documented for the September weights, experimental for 1017), not llama.cpp).), the better of Windows and Linux for Strata, an NVMe SSD; for the image and video rows, Qwen-Image (a planning budget): base model, INT8, 512×512, batch one, no reference images; Kolibri-1: the official FP8 route at a 32K context; MiniMax H3: local H3-Base FL2VA recipes, assuming an NVMe drive; Nemotron 3.5: Bolt and Super VL on one card with the RAM given; FLUX 3 Image Klein: Black Forest Labs' int8 ComfyUI route on the card. "With caveats" means part of the model sits in system RAM (slower), an unmeasured peak, or a route the checker marks conditional. Speeds appear only where they were measured on this very card in this setup; official estimates and runs on other setups are left out (the full checkers show them, labeled). Pick this card in a checker for your exact setup.

What more RAM would change

Measured a local model on your own hardware, or tried and it failed? Send us the numbers: paste the benchmark output or log into the form (or email it). After checking it we add it to this page, labeled as a community report and shared under CC BY 4.0.

About the Radeon VII / Instinct MI50 16 GB

Radeon VII / Instinct MI50 16 GB: 16 GB HBM2 of video memory (AMD accelerator specifications (Instinct MI50 16GB) and AMD graphics specifications (Radeon VII)). For the large mixture-of-experts models here, system RAM matters as much as VRAM: the parts that do not fit on the card are kept in RAM.

This page is the 16 GB Radeon VII and Instinct MI50. AMD also lists an Instinct MI50 with 32 GB; the checkers do not offer that version, so its answers are not on this page.

FAQ

Can a Radeon VII / Instinct MI50 16 GB run Qwen3.8-Flash-Next in the Strata engine?

Only on an experimental path, with 64 GB of system RAM in RigCheck's quick setup: IQ3_S; Linux only; No measured speed for this hardware and setup yet; Fits by the docs' rule (experts + ~10 GB), with little room to spare: close the browser while it runs. Covered by the community "Older GPUs" guide (Linux). The published MI50 benchmark used two cards. Its gfx906 build failed from engine 0.1.44 to 0.2.0; 0.2.1 builds again but has not been run on a card (AMD_HIP.md). The full checker takes your exact setup (RAM, disk, operating system, processor) and gives the install steps.

Can a Radeon VII / Instinct MI50 16 GB run Qwen3.8-27B in llama.cpp?

With caveats, by RigCheck's capacity estimate, with 64 GB of system RAM in RigCheck's quick setup: Q8_0 · system RAM, CPU only; No measured speed for this hardware and setup yet; RigCheck capacity estimate, not a measured setup. The full checker takes your exact setup (context length, RAM, image input) and gives the command to start it.

Can a Radeon VII / Instinct MI50 16 GB run Gemma 4 (E2B, E4B, 12B, 26B-A4B, 31B) in llama.cpp?

With caveats, by RigCheck's capacity estimate, with 64 GB of system RAM in RigCheck's quick setup: Gemma 4 31B: fits with caveats (capacity estimate) (Q8_0, system RAM, CPU only); Gemma 4 26B-A4B: fits with caveats (capacity estimate) (BF16, system RAM, CPU only); Gemma 4 12B: fits with caveats (capacity estimate) (BF16, system RAM, CPU only); Gemma 4 E4B: fits with caveats (capacity estimate) (BF16, system RAM, CPU only); Gemma 4 E2B: fits with caveats (capacity estimate) (BF16, system RAM, CPU only); No measured speed for this hardware and setup yet; RigCheck capacity estimate, not a measured setup. The full checker takes your exact setup (context length, RAM, image input) and gives the command to start it.

Other graphics cards

RTX 4060 Ti 16 GB · RTX 3060 12 GB · RTX 4070 Ti · RTX 3090 · RTX 4090 · RTX 5090 · Every 16 GB card · All graphics cards · Any other card or Mac

Sources