Skip to content

Olmo 3 1025 7B VRAM requirements

How much VRAM Olmo 3 1025 7B needs
4-bit weights: 4.1 GBWith 8K context: 7.4 GBWith 32K context: 10.4 GB8-bit, 8K context: 10.5 GB
Parameters
7.3B
Longest context
64K
Released
Sep 2025
License
Apache 2.0
Strongest at
Not rated

How much VRAM Olmo 3 1025 7B needs

Memory Olmo 3 1025 7B needs at each precision and context length
PrecisionWeightsWith 8KWith 32KWith 64K
16-bit13.6 GB16.9 GB19.9 GB23.9 GB
8-bit7.2 GB10.5 GB13.5 GB17.5 GB
6-bit5.6 GB8.8 GB11.8 GB15.8 GB
5-bit4.8 GB8.1 GB11.1 GB15.1 GB
4-bit4.1 GB7.4 GB10.4 GB14.4 GB
3-bit3.3 GB6.6 GB9.6 GB13.6 GB
2-bit2.5 GB5.8 GB8.8 GB12.8 GB
1.6-bit1.4 GB4.7 GB7.7 GB11.7 GB

One conversation, 16-bit context cache, runtime overhead included.

GPUs and Macs that run Olmo 3 1025 7B

The best way Olmo 3 1025 7B runs on common GPUs, Macs and AI boxes
MachineBest fitContextSpeedFirst replyReport
RTX 40603-bit16K54 tok/s1.0 sPDF: Olmo 3 1025 7B on RTX 4060 report
RTX 50603-bit16K86 tok/s0.8 sPDF: Olmo 3 1025 7B on RTX 5060 report
RTX 30703-bit16K86 tok/s0.7 sPDF: Olmo 3 1025 7B on RTX 3070 report
RTX 3080 10 GB5-bit16K98 tok/s0.5 sPDF: Olmo 3 1025 7B on RTX 3080 10 GB report
RTX 3060 12 GB6-bit28K43 tok/s1.2 sPDF: Olmo 3 1025 7B on RTX 3060 12 GB report
Arc B5806-bit28K37 tok/s0.8 sPDF: Olmo 3 1025 7B on Arc B580 report
RTX 40706-bit28K59 tok/s0.5 sPDF: Olmo 3 1025 7B on RTX 4070 report
RTX 4070 Super6-bit28K59 tok/s0.4 sPDF: Olmo 3 1025 7B on RTX 4070 Super report
RTX 50706-bit28K77 tok/s0.5 sPDF: Olmo 3 1025 7B on RTX 5070 report
RTX 4060 Ti 16 GB8-bit44K27 tok/s0.7 sPDF: Olmo 3 1025 7B on RTX 4060 Ti 16 GB report
RTX 5060 Ti 16 GB8-bit44K41 tok/s0.6 sPDF: Olmo 3 1025 7B on RTX 5060 Ti 16 GB report
Arc A770 16 GB8-bit44K36 tok/s0.7 sPDF: Olmo 3 1025 7B on Arc A770 16 GB report
Radeon RX 9070 XT8-bit44K48 tok/s0.4 sPDF: Olmo 3 1025 7B on Radeon RX 9070 XT report
RTX 4070 Ti Super8-bit44K61 tok/s0.3 sPDF: Olmo 3 1025 7B on RTX 4070 Ti Super report
RTX 4080 Super8-bit44K66 tok/s0.3 sPDF: Olmo 3 1025 7B on RTX 4080 Super report
RTX 5070 Ti8-bit44K79 tok/s0.3 sPDF: Olmo 3 1025 7B on RTX 5070 Ti report
RTX 50808-bit44K84 tok/s0.3 sPDF: Olmo 3 1025 7B on RTX 5080 report
Radeon RX 7900 XTX8-bit64K70 tok/s0.7 sPDF: Olmo 3 1025 7B on Radeon RX 7900 XTX report
RTX 30908-bit64K82 tok/s0.4 sPDF: Olmo 3 1025 7B on RTX 3090 report
RTX 40908-bit64K88 tok/s0.2 sPDF: Olmo 3 1025 7B on RTX 4090 report
Radeon AI PRO R97008-bit64K48 tok/s0.4 sPDF: Olmo 3 1025 7B on Radeon AI PRO R9700 report
RTX 50908-bit64K147 tok/s0.1 sPDF: Olmo 3 1025 7B on RTX 5090 report
2× RTX 30908-bit64K82 tok/s0.2 sPDF: Olmo 3 1025 7B on 2× RTX 3090 report
2× RTX 40908-bit64K88 tok/s0.1 sPDF: Olmo 3 1025 7B on 2× RTX 4090 report
RTX PRO 5000 Blackwell 48 GB8-bit64K114 tok/s0.2 sPDF: Olmo 3 1025 7B on RTX PRO 5000 Blackwell 48 GB report
RTX 6000 Ada8-bit64K84 tok/s0.2 sPDF: Olmo 3 1025 7B on RTX 6000 Ada report
L40S8-bit64K77 tok/s0.3 sPDF: Olmo 3 1025 7B on L40S report
2× RTX 50908-bit64K147 tok/s0.1 sPDF: Olmo 3 1025 7B on 2× RTX 5090 report
4× RTX 30908-bit64K82 tok/s0.1 sPDF: Olmo 3 1025 7B on 4× RTX 3090 report
RTX PRO 6000 Blackwell8-bit64K147 tok/s0.1 sPDF: Olmo 3 1025 7B on RTX PRO 6000 Blackwell report
Mac mini M4 16 GB6-bit16K17 tok/s9.5 sPDF: Olmo 3 1025 7B on Mac mini M4 16 GB report
Mac mini M4 24 GB8-bit44K13 tok/s9.5 sPDF: Olmo 3 1025 7B on Mac mini M4 24 GB report
Mac mini M4 32 GB8-bit64K13 tok/s9.5 sPDF: Olmo 3 1025 7B on Mac mini M4 32 GB report
Mac mini M4 Pro 24 GB8-bit44K28 tok/s4.8 sPDF: Olmo 3 1025 7B on Mac mini M4 Pro 24 GB report
Mac mini M4 Pro 48 GB8-bit64K28 tok/s4.8 sPDF: Olmo 3 1025 7B on Mac mini M4 Pro 48 GB report
Mac mini M4 Pro 64 GB8-bit64K28 tok/s4.8 sPDF: Olmo 3 1025 7B on Mac mini M4 Pro 64 GB report
M4 Max 128 GB8-bit64K49 tok/s2.4 sPDF: Olmo 3 1025 7B on M4 Max 128 GB report
M5 Max 128 GB8-bit64K54 tok/s0.7 sPDF: Olmo 3 1025 7B on M5 Max 128 GB report
M3 Ultra 96 GB8-bit64K61 tok/s1.3 sPDF: Olmo 3 1025 7B on M3 Ultra 96 GB report
M3 Ultra 256 GB8-bit64K61 tok/s1.3 sPDF: Olmo 3 1025 7B on M3 Ultra 256 GB report
M3 Ultra 512 GB8-bit64K61 tok/s1.3 sPDF: Olmo 3 1025 7B on M3 Ultra 512 GB report
M5 Ultra 512 GB8-bit64K84 tok/s0.4 sPDF: Olmo 3 1025 7B on M5 Ultra 512 GB report
Ryzen AI Max+ 395 64 GB8-bit64K20 tok/s1.4 sPDF: Olmo 3 1025 7B on Ryzen AI Max+ 395 64 GB report
Ryzen AI Max+ 395 128 GB8-bit64K20 tok/s1.4 sPDF: Olmo 3 1025 7B on Ryzen AI Max+ 395 128 GB report
RTX Spark8-bit64K28 tok/s0.5 sPDF: Olmo 3 1025 7B on RTX Spark report
DGX Spark8-bit64K25 tok/s0.5 sPDF: Olmo 3 1025 7B on DGX Spark report
Dual DGX Spark8-bit64K25 tok/s0.3 sPDF: Olmo 3 1025 7B on Dual DGX Spark report
DGX Station GB3008-bit64K411 tok/s0.1 sPDF: Olmo 3 1025 7B on DGX Station GB300 report
2× RTX PRO 6000 Blackwell8-bit64K147 tok/s0.1 sPDF: Olmo 3 1025 7B on 2× RTX PRO 6000 Blackwell report
4× RTX PRO 6000 Blackwell8-bit64K147 tok/s0.1 sPDF: Olmo 3 1025 7B on 4× RTX PRO 6000 Blackwell report
8× RTX PRO 6000 Blackwell8-bit64K147 tok/s0.1 sPDF: Olmo 3 1025 7B on 8× RTX PRO 6000 Blackwell report
A100 80 GB SXM8-bit64K164 tok/s0.3 sPDF: Olmo 3 1025 7B on A100 80 GB SXM report
H100 SXM8-bit64K245 tok/s0.1 sPDF: Olmo 3 1025 7B on H100 SXM report
4× H200 NVL8-bit64K319 tok/s0.1 sPDF: Olmo 3 1025 7B on 4× H200 NVL report
8× H200 NVL8-bit64K319 tok/s0.1 sPDF: Olmo 3 1025 7B on 8× H200 NVL report

Best fit is the highest quality, up to 8-bit, that leaves 16K of context. First reply is the wait for a 2K prompt. Speeds are rough, plus or minus 25%.

How to run it

llama.cpp
llama-server -hf mradermacher/Olmo-3-1025-7B-i1-GGUF
vLLM
vllm serve allenai/Olmo-3-1025-7B
MLX (Mac)
mlx_lm.chat --model mlx-community/Olmo-3-1025-7B-4bit

Questions

How much VRAM does Olmo 3 1025 7B need?

Olmo 3 1025 7B needs 4.1 GB for its weights at 4-bit, 7.4 GB with an 8K context and 10.4 GB with 32K. At 8-bit, 10.5 GB.

What is the minimum VRAM for Olmo 3 1025 7B?

About 6.6 GB at 3-bit with an 8K context, the lowest precision we recommend for a model this size. The smallest machine here that runs it without spilling is the RTX 4060.

How big is Olmo 3 1025 7B?

7.3B parameters: 13.6 GB of weights at 16-bit, 4.1 GB at 4-bit.

Can Olmo 3 1025 7B run on an RTX 4090?

Yes. At 8-bit it holds 64K of context and writes about 88 tokens a second.

More models