FlexOlmo 7x7B 1T — Hardware Requirements & GPU Compatibility
Chat
FlexOlmo 7x7B 1T is a 33.3B-parameter open language model from Allen AI in the OLMo family. It supports a context window of up to 4,096 tokens. At BF16 it needs about 67.92 GB of VRAM — see which GPUs and Macs can run it below.
FlexOlmo 7x7B 1T runs on a 96 GB+ Mac or a workstation GPU
Best local pick → Mac Studio (M3 Ultra, 96GB) · BF16 · 69 GB · ~16 tok/s
Macs from 96 GB (BF16 · ~16 tok/s)
Apache 2.0
Not set up for it?
Reference GPU · RTX 4090 24 GB
Estimates, ±20% · how we estimate
Run FlexOlmo 7x7B 1T
One recommended way. Every other method, OS and quant is in the step-by-step guide.
Recommended
vllm serve allenai/FlexOlmo-7x7B-1T --max-model-len 4096 --tensor-parallel-size 8vLLM repo with its own quantisation
Download: Weights on Hugging Face
Other ways
each opens step-by-stepRun in cloud
Fits on RTX PRO 6000 (96 GB) (27 GB headroom) · BF16
- Generation speed
- ~30 tok/s
- generation speed
- Cost per 1M output tokens
- $14.27
- per 1M output tokens
How Much VRAM Does FlexOlmo 7x7B 1T Need?
Select a quantization level; the listed hardware adheres to it.
| Quantization | VRAM | +32K | Runs on | |
|---|---|---|---|---|
| 67.9 GB | — | 16-bit · bfloat — Brain floating point 16 — preferred for training | Runs on RTX PRO 5000 Blackwell 72 GB |
Selected BF16
Brain floating point 1669.0GBat 4K context
- Weights
- 66.6 GB
- KV cache
- 2.1 GB
- Overhead
- 0.3 GB
Server-class: ~69 GB at 4K context · one RTX PRO 5000 Blackwell 72 GB
Which Hardware Can Run FlexOlmo 7x7B 1T?
BF16 · 67.9 GBFlexOlmo 7x7B 1T (BF16) requires 67.9 GB of VRAM to load the model weights. For comfortable inference with headroom for KV cache and system overhead, 89+ GB is recommended. Using the full 4K context window can add up to 1.1 GB, bringing total usage to 69.0 GB. No consumer GPU has enough memory. 30 devices with unified memory can run FlexOlmo 7x7B 1T, including Mac Studio (M3 Ultra, 256GB), Mac Studio (M3 Ultra, 512GB), Mac Studio (M3 Ultra, 96GB).
96 GB
fits · 27.0 GB free · up to 4K ctx
3 GPUs · 2 Mac classes
RTX PRO 6000 Blackwell Max-Q Workstation Edition~33Simulate output on RTX PRO 6000 Blackwell Max-Q Workstation EditionRTX PRO 6000 Blackwell Workstation Edition~33Simulate output on RTX PRO 6000 Blackwell Workstation EditionRTX PRO 6000 Blackwell Server Edition~30Simulate output on RTX PRO 6000 Blackwell Server EditionM3 Ultra*96 GB~16Simulate output on Mac Studio (M3 Ultra, 96GB)M3 Max*96 GB~6Simulate output on MacBook Pro 14" M3 Max (96 GB)128–144 GB
fits · 59.0 GB free · up to 4K ctx
4 GPUs · 2 Mac classes · 4 device families
GH200 Grace Hopper Superchip~83Simulate output on GH200 Grace Hopper SuperchipH200 NVL~81Simulate output on H200 NVLH200 SXM~81Simulate output on H200 SXMM2 Ultra*128 GB~15Simulate output on Mac Pro M2 Ultra (128 GB)M5 Max128 GB~12Simulate output on Mac Studio M5 Max (128 GB)GB10~5Simulate output on ASUS Ascent GX10Jetson T5000 (Thor)~5Simulate output on NVIDIA Jetson AGX Thor Developer Kit
By VRAM tier · typical GPUs and Macs at BF16
Consumer GPUs (up to 32 GB) and Macs up to 64 GB: too small for this model (needs ≥ 69 GB)8 GB tier →12 GB tier →16 GB tier →24 GB tier →Macs →
Alternatives to FlexOlmo 7x7B 1T
OLMo family · 20 models →Similar models
6| Model | Score | VRAM | tok/s | Why it is similar |
|---|---|---|---|---|
| MiniCPM V 4.7 35B A3BOpenBMB | — | 22.1 GB | ~169 | same size |
| Llm Jp 4.1 32B A3b Thinkingllm-jp | — | 20.1 GB | ~188 | same size |
| Llm Jp 4.1 33B Thinkingllm-jp | — | 22.5 GB | ~35 | same size |
| Agnes 3.0 FlashAgnes-AI | — | 22.1 GB | ~35 | same size |
| NVIDIA Nemotron 3.5 Lightning 30B A3B BF16NVIDIA | — | 19.3 GB | ~197 | same size |
| Agents A1InternScience | — | 21.5 GB | ~171 | same size |
Specifications
- Publisher
- Allen AI
- Family
- OLMo
- Parameters
- 33.3B
- Active parameters
- 16.1B per token
- Architecture
- FlexOlmoForCausalLM
- Context Length
- 4,096 tokens
- Vocabulary Size
- 100,352
- Release Date
- 2025-06-11
- License
- Apache 2.0
Frequently Asked Questions
How much VRAM does FlexOlmo 7x7B 1T need?
FlexOlmo 7x7B 1T requires 67.9 GB of VRAM at BF16. Full 4K context adds up to 1.1 GB (69.0 GB total).
VRAM = Weights + KV Cache + Overhead
Weights = 33.3B × 16 bits ÷ 8 = 66.5 GB
KV Cache + Overhead ≈ 1.4 GB (at 2K context + ~0.3 GB framework)
Fit ratings and hardware model lists check this model with room for a 16K-token context, which needs a little more memory.
KV Cache + Overhead ≈ 2.5 GB (at full 4K context)
VRAM usage by quantization
Can NVIDIA GeForce RTX 5090 run FlexOlmo 7x7B 1T?
No — FlexOlmo 7x7B 1T requires at least 67.9 GB at BF16, which exceeds the NVIDIA GeForce RTX 5090's 32 GB of VRAM.
Can I run FlexOlmo 7x7B 1T on a Mac?
Yes — Mac Studio (M3 Ultra, 96GB) and 20 other Macs can run FlexOlmo 7x7B 1T. Apple Silicon uses unified memory, so the model shares RAM with the system. At BF16 you need at least 67.9 GB of usable unified memory (RAM minus macOS overhead).
Can I run FlexOlmo 7x7B 1T locally?
Yes — FlexOlmo 7x7B 1T can run locally on consumer hardware. Popular tools include Ollama, LM Studio, and llama.cpp.
How fast is FlexOlmo 7x7B 1T?
At BF16, FlexOlmo 7x7B 1T can reach ~76 tok/s on AMD Instinct MI350X. Speed depends mainly on GPU memory bandwidth. Real-world results typically within ±20%.
tok/s = 1000 ÷ (active GB ÷ (bandwidth GB/s × efficiency) × 1000 + layers × routing ms)
Mixture-of-Experts: just the active experts get read for each token, along with a fixed routing cost per layer.
Example: NVIDIA B200 → 32.9 GB active ÷ (8000 × 0.65) = 6.32 ms, plus 32 layers × 0.055 ms = 1.76 ms, so 1000 ÷ 8.08 ms = ~124 tok/s
Estimated speed at BF16 (67.9 GB)
Real-world results typically within ±20%. Speed depends on batch size, quantization kernel, and software stack.
What's the download size of FlexOlmo 7x7B 1T?
At BF16, the download is about 66.54 GB.
Which GPUs can run FlexOlmo 7x7B 1T?
No single consumer GPU has enough VRAM to run FlexOlmo 7x7B 1T at BF16 (67.9 GB). Multi-GPU or professional hardware is required.
Which devices can run FlexOlmo 7x7B 1T?
30 devices with unified memory can run FlexOlmo 7x7B 1T at BF16 (67.9 GB), including ASUS Ascent GX10, Beelink GTR9 Pro (Ryzen AI Max+ 395, 128 GB), Framework Desktop (Ryzen AI Max+ 395, 128 GB), GMKtec EVO-X2 (Ryzen AI Max+ 395, 128 GB). Apple Silicon Macs use unified memory shared between CPU and GPU, making them well-suited for local LLM inference.