Calculate the TPS of the Radeon Pro Vega 48 on local AI models
Every model in our catalogue assessed against this card at the context length and minimum quality you choose. Speed is an estimate for a single request, calculated from this card's memory bandwidth and the size of each model once compressed.
Calculated for this card
679 models in our catalogue altogether
Largest model it holds
Baichuan 1-13B
13.3B · Q3_K_M · 27.0 tok/s
Fastest model
Gemma 3 QAT 1B
133 tok/s · 1B
Which AI models can run on a Radeon Pro Vega 48?
Set the inputs, read the answer
More context means more memory for the conversation cache. Speed is for a fresh conversation and does not change with this setting.
Hides models that would only fit by being compressed below this point.
337 models match
Calculating| Quantisation | Fit | ||||||
|---|---|---|---|---|---|---|---|
|
133
tok/s
80–213 · low confidence |
Gemma 3 1B | 1B | Mar 2025 | 1.8 GB | 33k tokens | Q8_0 | Comfortable |
|
133
tok/s
80–213 · low confidence |
Gemma 3 QAT 1B | 1B | Apr 2025 | 1.8 GB | 33k tokens | Q8_0 | Comfortable |
|
133
tok/s
80–213 · low confidence |
HGRN 1B (WT 103) ≈ | 1B | Nov 2023 | 1.8 GB | 131k tokens ? | Q8_0 | Comfortable |
|
133
tok/s
80–213 · low confidence |
LLama 3..2 Typhoon 2 1B ≈ | 1B | Dec 2024 | 1.8 GB | 131k tokens ? | Q8_0 | Comfortable |
|
133
tok/s
80–213 · low confidence |
OLMo-1B ≈ | 1B | Feb 2024 | 1.8 GB | 131k tokens ? | Q8_0 | Comfortable |
|
133
tok/s
80–213 · low confidence |
Pythia-1b ≈ | 1B | Apr 2023 | 1.8 GB | 131k tokens ? | Q8_0 | Comfortable |
|
123
tok/s
74–197 · low confidence |
OpenELM-1.1B ≈ | 1.1B | May 2024 | 1.9 GB | 131k tokens ? | Q8_0 | Comfortable |
|
121
tok/s
73–193 · low confidence |
DeciCoder-1B ≈ | 1.1B | Aug 2023 | 1.9 GB | 131k tokens ? | Q8_0 | Comfortable |
|
121
tok/s
73–193 · low confidence |
SantaCoder ≈ | 1.1B | Jan 2023 | 1.9 GB | 131k tokens ? | Q8_0 | Comfortable |
|
121
tok/s
73–193 · low confidence |
TinyLlama-1.1B (1T token checkpoint) ≈ | 1.1B | Oct 2023 | 1.9 GB | 131k tokens ? | Q8_0 | Comfortable |
|
121
tok/s
73–193 · low confidence |
TinyLlama-1.1B (3T token checkpoint) ≈ | 1.1B | Oct 2023 | 1.9 GB | 131k tokens ? | Q8_0 | Comfortable |
|
111
tok/s
66–177 · low confidence |
EXAONE 4.0 (1.2B) ≈ | 1.2B | Jul 2025 | 2.0 GB | 131k tokens ? | Q8_0 | Comfortable |
|
111
tok/s
66–177 · low confidence |
MinerU2.5 ≈ | 1.2B | Sep 2025 | 2.0 GB | 131k tokens ? | Q8_0 | Comfortable |
|
111
tok/s
66–177 · low confidence |
Pleias 1.0 1.2B ≈ | 1.2B | Dec 2024 | 2.0 GB | 131k tokens ? | Q8_0 | Comfortable |
|
111
tok/s
66–177 · low confidence |
Pleias-RAG-1B ≈ | 1.2B | Apr 2025 | 2.0 GB | 131k tokens ? | Q8_0 | Comfortable |
|
108
tok/s
65–173 · low confidence |
Llama 3.2 1B | 1.2B | Sep 2024 | 2.2 GB | 131k tokens | Q8_0 | Comfortable |
|
107
tok/s
64–171 · low confidence |
MiniCPM-1.2B ≈ | 1.2B | Jun 2024 | 2.0 GB | 131k tokens ? | Q8_0 | Comfortable |
|
102
tok/s
61–164 · low confidence |
DeepSeek Coder 1.3B ≈ | 1.3B | Jan 2024 | 2.1 GB | 131k tokens ? | Q8_0 | Comfortable |
|
102
tok/s
61–164 · low confidence |
DeepSeek-VL-1.3B ≈ | 1.3B | Mar 2024 | 2.1 GB | 131k tokens ? | Q8_0 | Comfortable |
|
102
tok/s
61–164 · low confidence |
DigiRL ≈ | 1.3B | Jun 2024 | 2.1 GB | 131k tokens ? | Q8_0 | Comfortable |
|
102
tok/s
61–164 · low confidence |
GLA Transformer 1.3B ≈ | 1.3B | Aug 2024 | 2.1 GB | 131k tokens ? | Q8_0 | Comfortable |
|
102
tok/s
61–164 · low confidence |
Janus 1.3B ≈ | 1.3B | Oct 2024 | 2.1 GB | 131k tokens ? | Q8_0 | Comfortable |
|
102
tok/s
61–164 · low confidence |
Kosmos-2.5 ≈ | 1.3B | Aug 2024 | 2.1 GB | 131k tokens ? | Q8_0 | Comfortable |
|
102
tok/s
61–164 · low confidence |
Otter ≈ | 1.3B | May 2023 | 2.1 GB | 131k tokens ? | Q8_0 | Comfortable |
|
102
tok/s
61–164 · low confidence |
Phi-1 ≈ | 1.3B | Oct 2023 | 2.1 GB | 131k tokens ? | Q8_0 | Comfortable |
Speeds are estimates for a single request — one conversation at a time — calculated from memory bandwidth, model size and quantisation. Real throughput varies with the inference runtime and its version. Figures published by hardware vendors measure many simultaneous requests and are much higher.
On record
Radeon Pro Vega 48 full specification
Everything on record for this board, ordered by how much it bears on running a language model rather than by how a spec sheet would list it. Memory comes first because it decides the outcome; the rest is context.
Memory
The two specifications that decide what this card can run and how quickly. Capacity sets which models fit; bandwidth sets how many tokens per second they produce once they do.
- Memory size
- 8 GB
- Memory bandwidth
- 402 GB/s
- Memory type
- HBM2
- Memory bus width
- 2,048 bit
- Memory clock
- 786 MHz
The chip
Which processor is on the board and how it was manufactured. A smaller process size generally means more performance for the same power.
- Graphics processor
- Vega 10
- Architecture
- GCN 5.0
- Generation
- Radeon Pro Mac(Vega Series)
- Foundry
- GlobalFoundries
- Process size
- 14 nm
- Transistors
- 12.5 billion
- Transistor density
- 25,300 K/mm²
- Die size
- 495 mm²
- Package
- BGA-2013
- Released
- 19 March 2019
Clock speeds
How fast the processor runs. Worth far less here than on a gaming benchmark: generating text is limited by memory bandwidth, so a higher clock barely moves the result.
- Base clock
- 1.2 GHz
- Boost clock
- 1.2 GHz
Processing units
What the chip contains. These drive graphics performance and matter mainly for processing a long prompt rather than for producing the answer.
- Shading units
- 3,072
- Texture mapping units
- 192
- Render output units
- 64
- L1 cache
- 16 KB
- L2 cache
- 4 MB
Theoretical performance
Peak arithmetic rates published for the board. These are ceilings that no real workload reaches, and generating text reaches a small fraction of them because it is limited by memory rather than arithmetic.
- Half precision (FP16)
- 14.8 TFLOPS
- Single precision (FP32)
- 7.4 TFLOPS
- Double precision (FP64)
- 460.8 GFLOPS
- Pixel rate
- 77 GPixel/s
- Texture rate
- 230 GTexel/s
The board
What it takes to physically install and power the card — the practical constraints that decide whether it fits the machine you already own.
- Power connectors
- None
- Bus interface
- PCIe 3.0 x16
- Slot width
- IGP
Software support
Which graphics and compute interfaces the card supports. CUDA compute capability is the one that bears on inference: below 7.0 there are no tensor cores, and modern inference software falls back to slower code paths.
- DirectX
- 12.1
- OpenGL
- 4.6
- Vulkan
- 1.3
- OpenCL
- 2.1
- Shader model
- 6.7
Listings
Where to buy a Radeon Pro Vega 48
No vendor is currently listing this card. Listings come from vendors who publish them here directly — browse the vendor directory to see who is selling what.
What the numbers mean
Why memory is the number that matters here
Memory
8 GB
Bandwidth
402 GB/s
Largest model
Baichuan 1-13B
Radeon Pro Vega 48 carries only 8 GB of HBM2. That limits it to the smaller end of the catalogue, and a model has to fit entirely inside before it generates anything at all. A runtime actually gets about 7.2 GB.
Memory bandwidth reaches 402 GB/s across a bus of 2,048 bits. That is the number governing generation speed — arithmetic per byte read is small enough that the bus, not the cores, is what everything waits on.
That comes from a memory clock of 786 MHz. Widening the bus and raising the clock are the two levers a manufacturer has, which is why a card with unremarkable cores can still generate quickly.
The biggest thing it holds is Baichuan 1-13B, 13.3B, compressed to Q3_K_M and generating around 27.0 tokens per second.
The chip and how it was built
Radeon Pro Vega 48 is built on the graphics processor Vega 10, using the architecture GCN 5.0 from AMD, as part of the generation Radeon Pro Mac(Vega Series).
The chip is manufactured by GlobalFoundries, on a process of 14 nm, with a die measuring 495 mm², holding 12.5 billion transistors. A smaller process generally means more performance for the same power, though for language models it matters far less than the memory subsystem.
It was released in March 2019, roughly 7.369551559902 years ago. Inference software support tends to follow hardware by a year or two, so a card of this age generally has mature, well-optimised code paths available to it.
Compute throughput, and why it matters less than it looks
FP16
14.8 TFLOPS
FP64
460.8 GFLOPS
On paper Radeon Pro Vega 48 reaches 14.8 TFLOPS at half precision, and 7.4 TFLOPS at single precision. These are peak figures no real workload sustains, and generating text reaches only a small fraction of them — decoding is limited by memory rather than arithmetic, which is why a card can look enormously powerful here and still produce tokens at an ordinary rate.
Double-precision throughput reaches 460.8 GFLOPS. It has no bearing on running a language model — no inference runtime uses it — but it separates datacentre parts from consumer ones, since the latter deliberately restrict it.
Clocks run from a base of 1.2 GHz to a boost of 1.2 GHz. Worth far less here than on a gaming benchmark: raising the clock speeds up the arithmetic, and the arithmetic is not what generation is waiting on.
Cache and processing units
Radeon Pro Vega 48 has an L1 cache of 16 KB, backed by an L2 cache of 4 MB. Cache absorbs a share of the memory traffic that would otherwise hit the main bus, which is the one place on this page where a number other than bandwidth quietly affects generation speed — a large L2 lets more of the working set stay close to the cores.
There are 3,072 shading units, 192 texture mapping units, and 64 render output units. These drive graphics workloads and contribute to prompt processing, but they sit idle for much of the time a model spends generating a reply.
Power, size and installation
The board occupies igp. Worth checking against the case and power supply already in the machine, since the largest cards need considerably more of both than a typical desktop provides.
It connects over PCIe 3.0 x16. The interface governs how quickly a model is loaded from disk into the card, not how fast it runs once there, so a narrower link costs a few seconds at startup and nothing thereafter.
The extremes
The largest AI models that run on a Radeon Pro Vega 48
The biggest open-weight models that fit on this card, newest first. Each is shown at the best compression the card can hold.
The fastest AI models on a Radeon Pro Vega 48
Where this card produces tokens quickest. Smaller models dominate here, because generating each token means reading the whole model out of memory once.
Step by step
How to work out the tokens per second of a Radeon Pro Vega 48
You do not have to calculate anything by hand — the gputps.com calculator on this page has already worked it out for every model this card can hold. Reading off the answer takes six steps.
-
01
Start with the model, not the specification
The table lists 337 models this card can run. Search by name, or by size — typing 27b matches on the parameter count even when the name never states it.
-
02
Match the context to your work
Drag the slider to the conversation length you plan to work at. The cache grows with the conversation, and against a card holding 8 GB that is frequently the difference between a model fitting and not.
-
03
Choose how far you will compress
By default the table picks the least-compressed copy that fits. Setting a floor removes models that only qualify through heavy compression.
-
04
Read the speed and the range
The figures are calculated, not measured. The fastest result on this card is 133 tok/s on Gemma 3 QAT 1B. The same card and model vary by thirty to fifty per cent between inference runtimes.
-
05
Read the fit verdict last
The fit column separates models that just fit from those with room to spare — worth checking before settling on one, against an available 8 GB.
-
06
Check the same model from the other side
Following a model through to its own page lists all the hardware that can run it, so you can see how it compares against Radeon Pro Vega 48.
Answers
Radeon Pro Vega 48 — common questions
Radeon Pro Vega 48— what is the largest AI model it can run?
The largest model in our catalogue that fits is Baichuan 1-13B at 13.3B parameters, compressed to Q3_K_M. It generates roughly 27.0 tokens per second and needs about 7.2 GB of the card's memory.
Radeon Pro Vega 48— how many tokens per second does it produce?
It depends on the model. The fastest model we track here is Gemma 3 QAT 1B at about 133 tokens per second, while larger models run proportionally slower because each token requires reading the whole model out of memory once. Speeds are estimates for a single conversation at a time.
Radeon Pro Vega 48— can it run 7B models?
Yes. For example it runs MetaMath 7B (LLaMa finetune) at Q4_K_M, using about 6.5 GB of memory and generating around 43.8 tokens per second.
Radeon Pro Vega 48— can it run 13B models?
Yes. For example it runs Gemma 4 12B at Q3_K_M, using about 6.5 GB of memory and generating around 30.0 tokens per second.
Radeon Pro Vega 48— how much memory does it have?
This card has 8 GB of HBM2. Around a tenth is reserved by the inference runtime and the driver, leaving roughly 7.2 GB available for a model and its conversation.
Radeon Pro Vega 48— what is its memory bandwidth?
Memory bandwidth reaches 402 GB/s across a bus of 2,048 bits. This is the single best predictor of how fast it generates text, because producing each token means reading the entire model out of memory once.
Radeon Pro Vega 48— what type of memory does it use?
It uses HBM2 clocked at 786 MHz. HBM types are found on datacentre accelerators and carry far more bandwidth than the GDDR used on desktop cards, which is why they generate tokens considerably faster at the same capacity.
Radeon Pro Vega 48— who makes it?
This is a product of AMD, with the chip manufactured by GlobalFoundries, on a process of 14 nm.
Radeon Pro Vega 48— when was it released?
It was released in March 2019.
Radeon Pro Vega 48— how much cache does it have?
The L1 cache is 16 KB, and the L2 cache is 4 MB. Cache absorbs part of the memory traffic that would otherwise reach the main bus, so a larger L2 gives a modest lift to generation speed beyond what bandwidth alone predicts.
Radeon Pro Vega 48— what are its TFLOPS?
It is rated at 14.8 TFLOPS at half precision and 7.4 TFLOPS at single precision. These are peak arithmetic ceilings rather than achievable rates, and text generation reaches only a small fraction of them because it is limited by memory bandwidth instead.
Radeon Pro Vega 48— does it support CUDA?
No. CUDA is NVIDIA-only, and this is a card from AMD. It runs language models through ROCm, Vulkan or Metal depending on the software, which are less mature than the CUDA path — our estimates apply a penalty for that.
Radeon Pro Vega 48— what bus interface does it use?
It uses PCIe 3.0 x16. This governs how fast a model is loaded onto the card rather than how fast it runs once loaded, so it costs a few seconds at startup and nothing during generation.
Radeon Pro Vega 48— is it good for running local AI models?
Its memory limits it to smaller models and its bandwidth gives usable, if unspectacular, generation speeds. In total it runs 337 of the models we track. Whether that is enough depends entirely on which model you want — the table above answers that directly.
Radeon Pro Vega 48— can it run a model that does not fit in its memory?
Offloading past the card's 8 GB is possible and usually a false economy: the system-memory portion is slow enough to dominate the result.
Would two Radeon Pro Vega 48 cards be twice as fast?
No. A second card doubles the memory to 16 GB of combined memory, at roughly the same generation speed as one.
Radeon Pro Vega 48— which AI models can it run?
337 of the 679 open-weight language models we track fit on this card and can be run locally. The table on this page lists every one, with the memory it needs, the quantisation it runs at and an estimated generation speed.
The other direction
Looking at it from the other side?
This page starts from the hardware. If you already know which model you want and need to know what it takes to run it, start from the model instead.