Calculate the TPS of the Radeon Instinct MI300X 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
Largest model it holds
Llama 4 Maverick
400B · Q3_K_M · 65.9 tok/s
Fastest model
Gemma 3 QAT 1B
1,759 tok/s · 1B
What AI models can a Radeon Instinct MI300X run?
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.
636 models match
Calculating| Quantisation | Fit | ||||||
|---|---|---|---|---|---|---|---|
|
1,759
tok/s
1,055–2,815 · low confidence |
Gemma 3 1B | 1B | Mar 2025 | 1.8 GB | 33k tokens | Q8_0 | Comfortable |
|
1,759
tok/s
1,055–2,815 · low confidence |
Gemma 3 QAT 1B | 1B | Apr 2025 | 1.8 GB | 33k tokens | Q8_0 | Comfortable |
|
1,759
tok/s
1,055–2,815 · low confidence |
HGRN 1B (WT 103) ≈ | 1B | Nov 2023 | 1.8 GB | 131k tokens ? | Q8_0 | Comfortable |
|
1,759
tok/s
1,055–2,815 · low confidence |
LLama 3..2 Typhoon 2 1B ≈ | 1B | Dec 2024 | 1.8 GB | 131k tokens ? | Q8_0 | Comfortable |
|
1,759
tok/s
1,055–2,815 · low confidence |
OLMo-1B ≈ | 1B | Feb 2024 | 1.8 GB | 131k tokens ? | Q8_0 | Comfortable |
|
1,759
tok/s
1,055–2,815 · low confidence |
Pythia-1b ≈ | 1B | Apr 2023 | 1.8 GB | 131k tokens ? | Q8_0 | Comfortable |
|
1,629
tok/s
977–2,606 · low confidence |
OpenELM-1.1B ≈ | 1.1B | May 2024 | 1.9 GB | 131k tokens ? | Q8_0 | Comfortable |
|
1,599
tok/s
960–2,559 · low confidence |
DeciCoder-1B ≈ | 1.1B | Aug 2023 | 1.9 GB | 131k tokens ? | Q8_0 | Comfortable |
|
1,599
tok/s
960–2,559 · low confidence |
SantaCoder ≈ | 1.1B | Jan 2023 | 1.9 GB | 131k tokens ? | Q8_0 | Comfortable |
|
1,599
tok/s
960–2,559 · low confidence |
TinyLlama-1.1B (1T token checkpoint) ≈ | 1.1B | Oct 2023 | 1.9 GB | 131k tokens ? | Q8_0 | Comfortable |
|
1,599
tok/s
960–2,559 · low confidence |
TinyLlama-1.1B (3T token checkpoint) ≈ | 1.1B | Oct 2023 | 1.9 GB | 131k tokens ? | Q8_0 | Comfortable |
|
1,466
tok/s
880–2,346 · low confidence |
EXAONE 4.0 (1.2B) ≈ | 1.2B | Jul 2025 | 2.0 GB | 131k tokens ? | Q8_0 | Comfortable |
|
1,466
tok/s
880–2,346 · low confidence |
MinerU2.5 ≈ | 1.2B | Sep 2025 | 2.0 GB | 131k tokens ? | Q8_0 | Comfortable |
|
1,466
tok/s
880–2,346 · low confidence |
Pleias 1.0 1.2B ≈ | 1.2B | Dec 2024 | 2.0 GB | 131k tokens ? | Q8_0 | Comfortable |
|
1,466
tok/s
880–2,346 · low confidence |
Pleias-RAG-1B ≈ | 1.2B | Apr 2025 | 2.0 GB | 131k tokens ? | Q8_0 | Comfortable |
|
1,430
tok/s
858–2,288 · low confidence |
Llama 3.2 1B | 1.2B | Sep 2024 | 2.2 GB | 131k tokens | Q8_0 | Comfortable |
|
1,410
tok/s
846–2,256 · low confidence |
MiniCPM-1.2B ≈ | 1.2B | Jun 2024 | 2.0 GB | 131k tokens ? | Q8_0 | Comfortable |
|
1,353
tok/s
812–2,165 · low confidence |
DeepSeek Coder 1.3B ≈ | 1.3B | Jan 2024 | 2.1 GB | 131k tokens ? | Q8_0 | Comfortable |
|
1,353
tok/s
812–2,165 · low confidence |
DeepSeek-VL-1.3B ≈ | 1.3B | Mar 2024 | 2.1 GB | 131k tokens ? | Q8_0 | Comfortable |
|
1,353
tok/s
812–2,165 · low confidence |
DigiRL ≈ | 1.3B | Jun 2024 | 2.1 GB | 131k tokens ? | Q8_0 | Comfortable |
|
1,353
tok/s
812–2,165 · low confidence |
GLA Transformer 1.3B ≈ | 1.3B | Aug 2024 | 2.1 GB | 131k tokens ? | Q8_0 | Comfortable |
|
1,353
tok/s
812–2,165 · low confidence |
Janus 1.3B ≈ | 1.3B | Oct 2024 | 2.1 GB | 131k tokens ? | Q8_0 | Comfortable |
|
1,353
tok/s
812–2,165 · low confidence |
Kosmos-2.5 ≈ | 1.3B | Aug 2024 | 2.1 GB | 131k tokens ? | Q8_0 | Comfortable |
|
1,353
tok/s
812–2,165 · low confidence |
Otter ≈ | 1.3B | May 2023 | 2.1 GB | 131k tokens ? | Q8_0 | Comfortable |
|
1,353
tok/s
812–2,165 · 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 Instinct MI300X 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
- 192 GB
- Memory bandwidth
- 5,325 GB/s
- Memory type
- HBM3
- Memory bus width
- 8,192 bit
- Memory clock
- 1.3 GHz
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
- Aqua Vanjaram
- Architecture
- CDNA 3.0
- Generation
- Radeon Instinct(MIx)
- Foundry
- TSMC
- Process size
- 5 nm
- Transistors
- 153 billion
- Transistor density
- 150,400 K/mm²
- Die size
- 1,017 mm²
- Package
- MCM
- Released
- 6 December 2023
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 GHz
- Boost clock
- 2.1 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
- 19,456
- Texture mapping units
- 1,216
- L1 cache
- 16 KB
- L2 cache
- 16 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)
- 653.7 TFLOPS
- Single precision (FP32)
- 81.7 TFLOPS
- Double precision (FP64)
- 81.7 TFLOPS
- Texture rate
- 2,554 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 draw (TDP)
- 750 W
- Suggested power supply
- 1,150 W
- Power connectors
- None
- Bus interface
- PCIe 5.0 x16
- Slot width
- OAM Module
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.
- OpenCL
- 3.0
Listings
Where to buy a Radeon Instinct MI300X
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
Capacity and bandwidth
Memory
192 GB
Bandwidth
5,325 GB/s
Largest model
Llama 4 Maverick
With 192 GB of HBM3, the Radeon Instinct MI300X is in the class of hardware that holds the largest open-weight models without splitting them across machines. Roughly 172.8 GB of that is reachable by an inference runtime once the driver takes its share.
Its 5,325 GB/s across a 8,192-bit bus is at the top of what exists. Since each token means reading the whole model out of memory once, that translates almost directly into generation speed — this card is bandwidth-rich enough that model size stops being the limiting factor long before the bus does.
The figure is the memory clock — 1.3 GHz here — multiplied by the bus width. It is why core counts predict generation speed so poorly.
The biggest thing it holds is Llama 4 Maverick (400B) at Q3_K_M compression, for about 65.9 tokens per second.
The chip and how it was built
The Radeon Instinct MI300X is built on the Aqua Vanjaram graphics processor, using AMD's CDNA 3.0 architecture, as part of the Radeon Instinct(MIx) generation.
The chip is manufactured by TSMC, on a 5 nm process, with a die measuring 1,017 mm², holding 153 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 December 2023, roughly 2 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
653.7 TFLOPS
FP64
81.7 TFLOPS
On paper the Radeon Instinct MI300X reaches 653.7 TFLOPS at half precision and 81.7 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 is 81.7 TFLOPS. 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 1 GHz at base to 2.1 GHz boosted. 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
The Radeon Instinct MI300X has 16 KB of L1 cache, backed by 16 MB of L2. 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 19,456 shading units, 1,216 texture mapping 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
Power draw
750 W
The Radeon Instinct MI300X is rated at 750 W, with a 1,150 W power supply suggested for the whole system. Running a language model keeps a card busy in bursts rather than continuously — it draws hard while generating and idles between requests — so sustained draw over a working day is usually well below the rated figure.
The board occupies a oam module. 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 5.0 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 a Radeon Instinct MI300X can run
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 Instinct MI300X
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 Instinct MI300X
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
Search for the model you want
The table lists 636 models this Radeon Instinct MI300X can run. Search by name, or by size — typing 27b matches on the parameter count even when the name never states it.
-
02
Set the context length you will actually use
Set the context to your real working length. Short questions cost almost nothing; a long document can consume a large share of the card's 192 GB.
-
03
Set a minimum quality if you need one
Compression is what lets bigger models fit. The quality control drops any model that needs more of it than you are willing to give.
-
04
Read the speed and the range
The figures are calculated, not measured. 1,759 tok/s on Gemma 3 QAT 1B is the fastest result on this card, and like every row it carries a range that reflects how much the runtime matters.
-
05
Check the headroom before you decide
Compare what each model needs with the 192 GB this card provides. Tight means it works today; comfortable means it still works when the conversation grows.
-
06
Open the model to compare cards
Following a model through to its own page lists all the hardware that can run it, so you can see where the Radeon Instinct MI300X sits against the alternatives.
Answers
Radeon Instinct MI300X — common questions
What bus interface does the Radeon Instinct MI300X use?
It uses PCIe 5.0 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.
Is the Radeon Instinct MI300X good for running local AI models?
Its memory is large enough for models most desktop hardware cannot touch and its bandwidth puts it among the fastest hardware available for generation. In total it runs 636 of the models we track. Whether that is enough depends entirely on which model you want — the table above answers that directly.
Can a Radeon Instinct MI300X run a model that does not fit in its memory?
Offloading past the card's 192 GB is possible and usually a false economy: the system-memory portion is slow enough to dominate the result.
Would two Radeon Instinct MI300X cards be twice as fast?
Capacity adds, throughput does not. Two of them give you 384 GB to work with rather than twice the tokens per second — every figure here is for a single Radeon Instinct MI300X.
What AI models can a Radeon Instinct MI300X run?
636 of the 679 open-weight language models we track fit on a Radeon Instinct MI300X and can be run locally on it. The table on this page lists every one, with the memory it needs, the quantisation it runs at and an estimated generation speed.
What is the largest AI model a Radeon Instinct MI300X can run?
The largest model in our catalogue that fits on a Radeon Instinct MI300X is Llama 4 Maverick at 400B parameters, compressed to Q3_K_M. It generates roughly 65.9 tokens per second and needs about 168.2 GB of the card's memory.
How many tokens per second does a Radeon Instinct MI300X produce?
It depends on the model. On a Radeon Instinct MI300X the fastest model we track is Gemma 3 QAT 1B at about 1,759 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.
Can a Radeon Instinct MI300X run a 7B model?
Yes. For example a Radeon Instinct MI300X runs Multi-Token Prediction 7B at Q8_0, using about 7.9 GB of memory and generating around 263 tokens per second.
Can a Radeon Instinct MI300X run a 13B model?
Yes. For example a Radeon Instinct MI300X runs DeepSeekMoE-16B at Q8_0, using about 17.5 GB of memory and generating around 611 tokens per second.
Can a Radeon Instinct MI300X run a 30B model?
Yes. For example a Radeon Instinct MI300X runs ERNIE-4.5-VL-28B-A3B at Q8_0, using about 29.2 GB of memory and generating around 349 tokens per second.
Can a Radeon Instinct MI300X run a 70B model?
Yes. For example a Radeon Instinct MI300X runs Qwen3-Coder-Next at Q8_0, using about 80.7 GB of memory and generating around 122 tokens per second.
How much memory does a Radeon Instinct MI300X have?
A Radeon Instinct MI300X has 192 GB of HBM3 memory. Around a tenth of that is reserved by the inference runtime and the driver, leaving roughly 172.8 GB available for a model and its conversation.
What is the memory bandwidth of a Radeon Instinct MI300X?
The Radeon Instinct MI300X has 5,325 GB/s of memory bandwidth, across a 8,192-bit memory bus. 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.
What type of memory does a Radeon Instinct MI300X use?
It uses HBM3 clocked at 1.3 GHz. 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.
Who makes the Radeon Instinct MI300X?
The Radeon Instinct MI300X is a AMD product, with the chip manufactured by TSMC, on a 5 nm process.
When was the Radeon Instinct MI300X released?
The Radeon Instinct MI300X was released in December 2023.
How much power does a Radeon Instinct MI300X use?
The Radeon Instinct MI300X has a rated board power of 750 W, and a 1,150 W system power supply is suggested. Generating text draws hard in bursts and idles between requests, so average consumption over a working session is normally well below the rated figure.
How much cache does a Radeon Instinct MI300X have?
The Radeon Instinct MI300X has 16 KB of L1 cache, and 16 MB of L2 cache. 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.
What are the TFLOPS of a Radeon Instinct MI300X?
The Radeon Instinct MI300X is rated at 653.7 TFLOPS at half precision and 81.7 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.
Does the Radeon Instinct MI300X support CUDA?
No. CUDA is NVIDIA-only, and the Radeon Instinct MI300X is a AMD card. 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.
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.