Skip to content

Sage A1.5 VRAM requirements

How much VRAM Sage A1.5 needs
4-bit weights: 20.3 GBWith 8K context: 21.5 GBWith 32K context: 22.0 GB8-bit, 8K context: 36.8 GB
Parameters
36B
Active per token
4.7B
Longest context
256K
Released
Oct 2026
License
Apache 2.0
Strongest at
Not rated

How much VRAM Sage A1.5 needs

Memory Sage A1.5 needs at each precision and context length
PrecisionWeightsWith 8KWith 32KWith 128K
16-bit67.0 GB68.2 GB68.6 GB70.5 GB
8-bit35.6 GB36.8 GB37.2 GB39.1 GB
6-bit27.5 GB28.7 GB29.1 GB31.0 GB
5-bit23.8 GB25.0 GB25.5 GB27.4 GB
4-bit20.3 GB21.5 GB22.0 GB23.8 GB
3-bit16.4 GB17.6 GB18.0 GB19.9 GB
2-bit12.6 GB13.8 GB14.2 GB16.1 GB
1.6-bit6.9 GB8.1 GB8.6 GB10.5 GB

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

GPUs and Macs that run Sage A1.5

The best way Sage A1.5 runs on common GPUs, Macs and AI boxes
MachineBest fitContextSpeedFirst replyReport
RTX 4060Too bigPDF: Sage A1.5 on RTX 4060 report
RTX 5060Too bigPDF: Sage A1.5 on RTX 5060 report
RTX 3070Too bigPDF: Sage A1.5 on RTX 3070 report
RTX 3080 10 GBToo bigPDF: Sage A1.5 on RTX 3080 10 GB report
RTX 3060 12 GBToo bigPDF: Sage A1.5 on RTX 3060 12 GB report
Arc B580Too bigPDF: Sage A1.5 on Arc B580 report
RTX 4070Too bigPDF: Sage A1.5 on RTX 4070 report
RTX 4070 SuperToo bigPDF: Sage A1.5 on RTX 4070 Super report
RTX 5070Too bigPDF: Sage A1.5 on RTX 5070 report
RTX 4060 Ti 16 GBToo bigPDF: Sage A1.5 on RTX 4060 Ti 16 GB report
RTX 5060 Ti 16 GBToo bigPDF: Sage A1.5 on RTX 5060 Ti 16 GB report
Arc A770 16 GBToo bigPDF: Sage A1.5 on Arc A770 16 GB report
Radeon RX 9070 XTToo bigPDF: Sage A1.5 on Radeon RX 9070 XT report
RTX 4070 Ti SuperToo bigPDF: Sage A1.5 on RTX 4070 Ti Super report
RTX 4080 SuperToo bigPDF: Sage A1.5 on RTX 4080 Super report
RTX 5070 TiToo bigPDF: Sage A1.5 on RTX 5070 Ti report
RTX 5080Too bigPDF: Sage A1.5 on RTX 5080 report
Radeon RX 7900 XTX4-bit72K115 tok/s1.2 sPDF: Sage A1.5 on Radeon RX 7900 XTX report
RTX 30904-bit72K145 tok/s0.8 sPDF: Sage A1.5 on RTX 3090 report
RTX 40904-bit72K152 tok/s0.3 sPDF: Sage A1.5 on RTX 4090 report
Radeon AI PRO R97006-bit96K72 tok/s0.8 sPDF: Sage A1.5 on Radeon AI PRO R9700 report
RTX 50906-bit96K180 tok/s0.3 sPDF: Sage A1.5 on RTX 5090 report
2× RTX 30908-bit256K98 tok/s0.5 sPDF: Sage A1.5 on 2× RTX 3090 report
2× RTX 40908-bit256K104 tok/s0.2 sPDF: Sage A1.5 on 2× RTX 4090 report
RTX PRO 5000 Blackwell 48 GB8-bit256K128 tok/s0.4 sPDF: Sage A1.5 on RTX PRO 5000 Blackwell 48 GB report
RTX 6000 Ada8-bit256K100 tok/s0.3 sPDF: Sage A1.5 on RTX 6000 Ada report
L40S8-bit256K92 tok/s0.5 sPDF: Sage A1.5 on L40S report
2× RTX 50908-bit256K155 tok/s0.2 sPDF: Sage A1.5 on 2× RTX 5090 report
4× RTX 30908-bit256K98 tok/s0.2 sPDF: Sage A1.5 on 4× RTX 3090 report
RTX PRO 6000 Blackwell8-bit256K155 tok/s0.2 sPDF: Sage A1.5 on RTX PRO 6000 Blackwell report
Mac mini M4 16 GBToo bigPDF: Sage A1.5 on Mac mini M4 16 GB report
Mac mini M4 24 GBToo bigPDF: Sage A1.5 on Mac mini M4 24 GB report
Mac mini M4 32 GB3-bit152K34 tok/s18 sPDF: Sage A1.5 on Mac mini M4 32 GB report
Mac mini M4 Pro 24 GBToo bigPDF: Sage A1.5 on Mac mini M4 Pro 24 GB report
Mac mini M4 Pro 48 GB6-bit256K43 tok/s8.9 sPDF: Sage A1.5 on Mac mini M4 Pro 48 GB report
Mac mini M4 Pro 64 GB8-bit256K35 tok/s8.9 sPDF: Sage A1.5 on Mac mini M4 Pro 64 GB report
M4 Max 128 GB8-bit256K55 tok/s4.4 sPDF: Sage A1.5 on M4 Max 128 GB report
M5 Max 128 GB8-bit256K60 tok/s1.3 sPDF: Sage A1.5 on M5 Max 128 GB report
M3 Ultra 96 GB8-bit256K66 tok/s2.5 sPDF: Sage A1.5 on M3 Ultra 96 GB report
M3 Ultra 256 GB8-bit256K66 tok/s2.5 sPDF: Sage A1.5 on M3 Ultra 256 GB report
M3 Ultra 512 GB8-bit256K66 tok/s2.5 sPDF: Sage A1.5 on M3 Ultra 512 GB report
M5 Ultra 512 GB8-bit256K82 tok/s0.8 sPDF: Sage A1.5 on M5 Ultra 512 GB report
Ryzen AI Max+ 395 64 GB8-bit256K27 tok/s2.6 sPDF: Sage A1.5 on Ryzen AI Max+ 395 64 GB report
Ryzen AI Max+ 395 128 GB8-bit256K27 tok/s2.6 sPDF: Sage A1.5 on Ryzen AI Max+ 395 128 GB report
RTX Spark8-bit256K37 tok/s0.9 sPDF: Sage A1.5 on RTX Spark report
DGX Spark8-bit256K34 tok/s0.9 sPDF: Sage A1.5 on DGX Spark report
Dual DGX Spark8-bit256K34 tok/s0.5 sPDF: Sage A1.5 on Dual DGX Spark report
DGX Station GB3008-bit256K292 tok/s0.1 sPDF: Sage A1.5 on DGX Station GB300 report
2× RTX PRO 6000 Blackwell8-bit256K155 tok/s0.1 sPDF: Sage A1.5 on 2× RTX PRO 6000 Blackwell report
4× RTX PRO 6000 Blackwell8-bit256K155 tok/s0.1 sPDF: Sage A1.5 on 4× RTX PRO 6000 Blackwell report
8× RTX PRO 6000 Blackwell8-bit256K155 tok/s0.1 sPDF: Sage A1.5 on 8× RTX PRO 6000 Blackwell report
A100 80 GB SXM8-bit256K167 tok/s0.6 sPDF: Sage A1.5 on A100 80 GB SXM report
H100 SXM8-bit256K218 tok/s0.2 sPDF: Sage A1.5 on H100 SXM report
4× H200 NVL8-bit256K255 tok/s0.1 sPDF: Sage A1.5 on 4× H200 NVL report
8× H200 NVL8-bit256K255 tok/s0.1 sPDF: Sage A1.5 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 InternScience/Sage-A1.5-Q4_K_M-GGUF
vLLM
vllm serve InternScience/Sage-A1.5
MLX (Mac)
mlx_lm.chat --model mlx-community/Sage-A1.5-OptiQ-4bit

Questions

How much VRAM does Sage A1.5 need?

Sage A1.5 needs 20.3 GB for its weights at 4-bit, 21.5 GB with an 8K context and 22.0 GB with 32K. At 8-bit, 36.8 GB.

What is the minimum VRAM for Sage A1.5?

About 17.6 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 Radeon RX 7900 XTX.

How big is Sage A1.5?

36B parameters (4.7B active per token): 67.0 GB of weights at 16-bit, 20.3 GB at 4-bit.

Can Sage A1.5 run on an RTX 4090?

Yes. At 4-bit it holds 72K of context and writes about 152 tokens a second.

More models