ether0 TPS calculator
Each card below is assessed against this model at the context length and minimum quality you choose. Speed is an estimate for a single request, calculated from the card's memory bandwidth and the size of the model once compressed.
Calculated for this model
818 cards we hold specifications for
Smallest card that fits
Xeon Phi 7120P
16 GB · IQ4_XS · 9.9 tok/s
Fastest card
B200
141 tok/s · 180 GB
Which GPUs can run ether0?
Set the inputs, read the answer
A longer conversation needs more memory, which can push this model off smaller cards.
Hides cards that would only fit the model by compressing it below this point.
241 cards match
Calculating| Needs | Quantisation | Fit | |||||
|---|---|---|---|---|---|---|---|
|
141
tok/s
85–226 · low confidence |
B200 NVIDIA | 180 GB | 8,000 GB/s | Jan 2024 | 26.4 GB | Q8_0 | Comfortable |
|
141
tok/s
85–226 · low confidence |
B300 NVIDIA | 288 GB | 8,000 GB/s | Sep 2025 | 26.4 GB | Q8_0 | Comfortable |
|
113
tok/s
68–180 · low confidence |
Radeon Instinct MI350X AMD | 288 GB | 8,190 GB/s | Jan 2025 | 26.4 GB | Q8_0 | Comfortable |
|
113
tok/s
68–180 · low confidence |
Radeon Instinct MI355X AMD | 288 GB | 8,190 GB/s | Jan 2025 | 26.4 GB | Q8_0 | Comfortable |
|
90.2
tok/s
54–144 · low confidence |
Radeon Instinct MI300 AMD | 128 GB | 6,550 GB/s | Jan 2023 | 26.4 GB | Q8_0 | Comfortable |
|
86.3
tok/s
52–138 · low confidence |
H200 NVL NVIDIA | 141 GB | 4,890 GB/s | Nov 2024 | 26.4 GB | Q8_0 | Comfortable |
|
86.3
tok/s
52–138 · low confidence |
H200 SXM 141 GB NVIDIA | 141 GB | 4,890 GB/s | Nov 2024 | 26.4 GB | Q8_0 | Comfortable |
|
82.6
tok/s
50–132 · low confidence |
Radeon Instinct MI325X AMD | 256 GB | 6,000 GB/s | Oct 2024 | 26.4 GB | Q8_0 | Comfortable |
|
73.3
tok/s
44–117 · low confidence |
Radeon Instinct MI300A AMD | 128 GB | 5,325 GB/s | Dec 2023 | 26.4 GB | Q8_0 | Comfortable |
|
73.3
tok/s
44–117 · low confidence |
Radeon Instinct MI300X AMD | 192 GB | 5,325 GB/s | Dec 2023 | 26.4 GB | Q8_0 | Comfortable |
|
73.3
tok/s
44–117 · low confidence |
Radeon Instinct MI308X AMD | 192 GB | 5,325 GB/s | Dec 2023 | 26.4 GB | Q8_0 | Comfortable |
|
69.5
tok/s
42–111 · low confidence |
H100 NVL 94 GB NVIDIA | 94 GB | 3,940 GB/s | Mar 2023 | 26.4 GB | Q8_0 | Comfortable |
|
59.3
tok/s
36–95 · low confidence |
H100 PCIe 96 GB NVIDIA | 96 GB | 3,360 GB/s | Mar 2023 | 26.4 GB | Q8_0 | Comfortable |
|
59.3
tok/s
36–95 · low confidence |
H100 SXM5 80 GB NVIDIA | 80 GB | 3,360 GB/s | Oct 2022 | 26.4 GB | Q8_0 | Comfortable |
|
59.3
tok/s
36–95 · low confidence |
H100 SXM5 94 GB NVIDIA | 94 GB | 3,360 GB/s | Mar 2023 | 26.4 GB | Q8_0 | Comfortable |
|
59.3
tok/s
36–95 · low confidence |
H100 SXM5 96 GB NVIDIA | 96 GB | 3,360 GB/s | Mar 2023 | 26.4 GB | Q8_0 | Comfortable |
|
59.3
tok/s
36–95 · low confidence |
H800 SXM5 NVIDIA | 80 GB | 3,360 GB/s | Mar 2023 | 26.4 GB | Q8_0 | Comfortable |
|
49.0
tok/s
29–78 · low confidence |
Tesla V100 SXM2 16 GB NVIDIA | 16 GB | 1,130 GB/s | Nov 2019 | 13.8 GB | IQ4_XS | Tight |
|
45.2
tok/s
27–72 · low confidence |
Radeon Instinct MI250 AMD | 128 GB | 3,280 GB/s | Nov 2021 | 26.4 GB | Q8_0 | Comfortable |
|
45.2
tok/s
27–72 · low confidence |
Radeon Instinct MI250X AMD | 128 GB | 3,280 GB/s | Nov 2021 | 26.4 GB | Q8_0 | Comfortable |
|
41.6
tok/s
25–67 · low confidence |
GeForce RTX 5080 NVIDIA | 16 GB | 960 GB/s | Jan 2025 | 13.8 GB | IQ4_XS | Tight |
|
38.9
tok/s
23–62 · low confidence |
Tesla V100 DGXS 16 GB NVIDIA | 16 GB | 897 GB/s | Mar 2018 | 13.8 GB | IQ4_XS | Tight |
|
38.9
tok/s
23–62 · low confidence |
Tesla V100 PCIe 16 GB NVIDIA | 16 GB | 897 GB/s | Jun 2017 | 13.8 GB | IQ4_XS | Tight |
|
38.9
tok/s
23–62 · low confidence |
Tesla V100 SXM2 16 GB NVIDIA | 16 GB | 897 GB/s | Jun 2017 | 13.8 GB | IQ4_XS | Tight |
|
38.8
tok/s
23–62 · low confidence |
GeForce RTX 5070 Ti NVIDIA | 16 GB | 896 GB/s | Feb 2025 | 13.8 GB | IQ4_XS | Tight |
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
Full specification
Everything on record for this model. Most of it describes how it was trained rather than how it runs — useful context for judging how much work went into it, and how it compares with models built at a different scale.
Origin
Who built this model, where, and when it was published.
- Organisation
- FutureHouse
- Country
- United States of America
- Published
- 5 June 2025
- Authors
- Siddharth Narayanan, James Braza, Albert Bou, Geemi Wellawatte, Mayk Caldas, Ludovico Mitchener, Sam Rodriques, Andrew White
What it does
The problem areas the model was built for. A model can carry several of each.
- Domain
- Materials science, Language
- Task
- Language modeling/generation, Molecular property prediction
- Base model
- Mistral Small 3
Size
How large the model is and how much data it was trained on. Parameters are the figure that decides whether it fits on a given graphics card.
- Parameters
- 24B
- Training data
- tokens
24B
Reasoning models are large language models that emit a long chain-of-thought before answering, providing both higher accuracy and explicit reasoning for their response. A major question has been whether language model reasoning generalizes beyond mathematics, programming, and logic, where most previous work has focused. We demonstrate that reasoning models can be post-trained for chemistry without additional domain pretraining, and require substantially less data compared to contemporary domain-…
Training compute
The arithmetic performed to train the model, measured in floating-point operations. It is a measure of what the training run cost, not of how fast the finished model answers you.
- How it was established
- Hardware
- Fine-tuning compute
- 3.9 × 10²² FLOP
989400000000000 FLOP/GPU/sec * 96 hours * 3600 sec / hour * 384 GPUs * 0.3 [assumed utilization] = 3.9391100928e+22 FLOP "Likely" confidence because I didn't account for SFT training, only RL
The training run
What it physically took to train: which chips, how many, for how long, and what that drew from the wall.
- Training hardware
- NVIDIA H100 SXM5 80GB
- Chips used
- 384
- Wall-clock time
- 96 hours
- Power draw
- 526.5 kW
"The seven specialist models were trained using 24-72 Nvidia H100 GPUs each" "The final all-task RL training phase was performed using 384 H100 GPUs, over 4 days" 4 days = 96 hours
Availability
Whether you can obtain the model and run it on your own hardware, which is what decides if any of the graphics-card figures on this page apply.
- Weights
- Open — downloadable
- Model access
- Open weights (unrestricted)
- Training code
- Unreleased
- Hugging Face
- futurehouse
apache 2.0 https://huggingface.co/futurehouse/ether0
How it is classified
Labels the source dataset applies when tracking notable models, and how confident it is in the entry.
- Record confidence
- Likely
Sources
Where this record came from and when it was last checked.
- Reference
- ether0: a scientific reasoning model for chemistry
- Last updated
- 28 November 2025
The extremes
The ten fastest GPUs that run ether0
Ranked by estimated tokens per second, newest card first where speeds tie. Because generation is bound by memory bandwidth, this ordering follows bandwidth rather than any gaming benchmark.
- 01 B300 288 GB · 8,000 GB/s · Q8_0 141 tok/s
- 02 B200 180 GB · 8,000 GB/s · Q8_0 141 tok/s
- 03 Radeon Instinct MI350X 288 GB · 8,190 GB/s · Q8_0 113 tok/s
- 04 Radeon Instinct MI355X 288 GB · 8,190 GB/s · Q8_0 113 tok/s
- 05 Radeon Instinct MI300 128 GB · 6,550 GB/s · Q8_0 90.2 tok/s
- 06 H200 NVL 141 GB · 4,890 GB/s · Q8_0 86.3 tok/s
- 07 H200 SXM 141 GB 141 GB · 4,890 GB/s · Q8_0 86.3 tok/s
- 08 Radeon Instinct MI325X 256 GB · 6,000 GB/s · Q8_0 82.6 tok/s
- 09 Radeon Instinct MI300A 128 GB · 5,325 GB/s · Q8_0 73.3 tok/s
- 10 Radeon Instinct MI300X 192 GB · 5,325 GB/s · Q8_0 73.3 tok/s
The smallest GPUs that still run ether0
The cheapest route in, by memory capacity. A tight fit runs the model but leaves nothing spare for a longer conversation.
- 01 Ryzen AI Z2 Extreme GPU 16 GB · needs 13.8 GB · IQ4_XS · tight 8.7 tok/s
- 02 Radeon RX 7700 16 GB · needs 13.8 GB · IQ4_XS · tight 21.0 tok/s
- 03 Arc Pro B50 16 GB · needs 13.8 GB · IQ4_XS · tight 6.3 tok/s
- 04 RTX PRO 2000 Blackwell 16 GB · needs 13.8 GB · IQ4_XS · tight 12.5 tok/s
- 05 Ryzen Z2 Extreme GPU 16 GB · needs 13.8 GB · IQ4_XS · tight 4.3 tok/s
- 06 Radeon RX 9060 XT 16 GB 16 GB · needs 13.8 GB · IQ4_XS · tight 10.9 tok/s
- 07 GeForce RTX 5060 Ti 16 GB 16 GB · needs 13.8 GB · IQ4_XS · tight 19.4 tok/s
- 08 GeForce RTX 5080 Mobile 16 GB · needs 13.8 GB · IQ4_XS · tight 38.8 tok/s
- 09 Radeon RX 9070 16 GB · needs 13.8 GB · IQ4_XS · tight 21.8 tok/s
- 10 Radeon RX 9070 XT 16 GB · needs 13.8 GB · IQ4_XS · tight 21.8 tok/s
What the numbers mean
Hardware requirements in practice
Minimum card
Xeon Phi 7120P
Memory needed
13.8 GB
Fastest
141 tok/s
With 24B parameters, ether0 lands in the range a serious desktop card can handle once the weights are compressed. 241 of the cards we track can run it.
The entry point is the Xeon Phi 7120P: 16 GB of memory, IQ4_XS compression, roughly 9.9 tokens per second.
The quickest result comes from a B200 at around 141 tokens per second — its 8,000 GB/s of bandwidth is what buys that.
What this model is
ether0 was published by FutureHouse, in United States of America, in June 2025.
It works in Materials science, Language, and is recorded as doing language modeling/generation, Molecular property prediction.
It is derived from Mistral Small 3 rather than trained from scratch, which is the usual way a specialised model is produced.
The weights are published, so it can be downloaded and run on your own hardware indefinitely, offline, with no account attached. It is published under the futurehouse organisation on Hugging Face.
What decides the speed
Half the cards that hold it manage more than 19.4 tokens per second, and 193 exceed reading speed outright.
It is a dense model, so every parameter is read for every token produced. That makes speed track memory bandwidth almost exactly — a card with twice the bandwidth generates roughly twice as fast.
Without the attention layout on record, the memory column is an approximation. It is close enough to choose hardware by, and least reliable at long context.
Step by step
How to choose a GPU for ether0
The table above has already assessed every card we hold specifications for against this model. Getting to your answer takes six steps.
-
01
Start from the memory column
The table lists every card that can hold ether0 — around 13.8 GB at IQ4_XS. That figure, not the card's headline performance, is what decides whether it runs.
-
02
Decide how long your conversations run
The conversation occupies memory too, and grows as it goes. Set the slider to the length you expect: at long context ether0 can slip off a card that handles short questions easily.
-
03
Set a quality floor
The quantisation column varies by card, because a bigger card holds a more accurate copy of ether0 — IQ4_XS on the smallest card that fits. Set a floor to hold the comparison at one level.
-
04
Sort by speed
Sort by speed to see how cards rank for ether0. It will not match a gaming ordering — generation is bound by memory bandwidth, which is why the B200 tops it at 141 tok/s.
-
05
Look at the headroom, not just the fit
A tight fit runs ether0 but leaves nothing spare for a longer conversation; comfortable has headroom. If you expect to grow the context, buy for comfortable.
-
06
Check the card from the other side
Each card page repeats this sweep for every model we hold. It answers what else the hardware is good for, beyond ether0.
Answers
ether0 — common questions
How accurate are these ether0 speed estimates?
These are estimates with real error bars. The fastest result here, 85–226 tok/s on the B200, could reasonably land anywhere in its published range depending on which runtime you use.
What GPU do I need to run ether0?
The smallest card in our catalogue that holds ether0 is the Xeon Phi 7120P, with 16 GB of memory. It runs the model at IQ4_XS using about 13.8 GB, and produces roughly 9.9 tokens per second. 241 cards in total can run it.
How fast is ether0 on a GPU?
It depends on the card. The quickest we calculate is a B200 at about 141 tokens per second; the slowest that still runs it manages considerably less. Reading speed is around ten tokens per second, and 193 of the cards that can run ether0 clear that.
How much VRAM does ether0 need?
About 13.8 GB at IQ4_XS compression, which is what the smallest card that runs it uses. Less compression needs more: the figures in the memory column above are recalculated for each card, because each one holds the least-compressed version it can.
Can I run ether0 on a 16 GB GPU?
Yes. A Tesla V100 SXM2 16 GB with 16 GB runs it at IQ4_XS, using about 13.8 GB and generating roughly 49.0 tokens per second — a tight fit.
Can I run ether0 on a 24 GB GPU?
Yes. A GeForce RTX 5090 D V2 with 24 GB runs it at Q6_K, using about 20.8 GB and generating roughly 34.4 tokens per second — a tight fit.
Is ether0 open source?
Its weights are published, so ether0 can be downloaded and run on your own hardware. Note that open weights is not the same as open source in the full sense — it says nothing about the training data, the training code, or the commercial terms attached.
How many parameters does ether0 have?
ether0 has 24B parameters. 24B. That figure is the total, and it is what decides how much memory the model needs — roughly half a gigabyte per billion at the compression most people use.
Who created ether0?
ether0 was published by FutureHouse, based in United States of America.
When was ether0 released?
ether0 was published in June 2025.
What is ether0 used for?
ether0 works in Materials science, Language, and is recorded as handling language modeling/generation, Molecular property prediction. Models frequently carry more than one of each, and the tags describe purpose rather than capability limits.
Where can I download ether0?
Its weights are published under the futurehouse organisation on Hugging Face. We do not host model files — this site calculates what hardware is needed to run them.
Can I run ether0 if it does not fit in my GPU?
It can be split between the card and system memory, but ether0 generates painfully slowly that way — the nearest miss we calculate is short by 4.4 GB. Nothing on this page assumes offloading.
Would two GPUs run ether0 faster?
Two cards buy memory rather than speed. That matters for ether0 only if one card cannot hold it — 241 can, so a second adds little.
Why does the quantisation differ between cards for ether0?
A larger card holds a more accurate copy. Across the cards that run ether0, 4 compression levels are used; the floor control above pins it to one.
The other direction
Looking at it from the other side?
This page starts from the model. If you already own a card and want to know everything it will run, start from the hardware instead.