RunsOnMy ModelsGPUs & MacsAPI pricesVRAM calculatorNew

Can I run Trinity Mini locally?

Trinity Mini by arcee-ai has 26B parameters (2.0B active per token, 128 experts, 8 used per token). At 4-bit with an 8K context it needs about 15.8 GB of GPU memory. It fits on 58 of 65 GPUs and Macs we track.

15.8 GBQ4_K_M · 8K context
RX 7900 XTsmallest consumer GPU that fits
M4 32GBsmallest Mac that fits
128Kmax context
GQA + sliding windowattention type

Memory needed by quant and context length

GB of GPU memory = weights + KV cache (FP16) + runtime overhead. Weights use real file sizes from arcee-ai/Trinity-Mini-GGUF where available (marked •), otherwise parameter count × bits per weight. How we calculate.

QuantQuality4K ctx8K ctx32K ctx128K ctx
2-bit (Q2_K) •rough9.7 GB9.8 GB10.2 GB11.7 GB
3-bit (Q3_K_M) •usable12.2 GB12.3 GB12.6 GB14.1 GB
4-bit (Q4_K_M) •good, most popular15.8 GB15.8 GB16.2 GB17.7 GB
5-bit (Q5_K_M) •very good18.3 GB18.4 GB18.7 GB20.2 GB
6-bit (Q6_K) •near lossless21.0 GB21.0 GB21.4 GB22.9 GB
8-bit (Q8_0) •effectively lossless26.8 GB26.9 GB27.3 GB28.8 GB
16-bit (F16/BF16)original weights49.6 GB49.7 GB50.0 GB51.5 GB

Which GPUs and Macs run Trinity Mini?

Best quality quant that fits at 8K context. "Tight" = within 10% of usable memory. Speed is a ceiling from memory bandwidth (real speed is typically 50–80% of it).

HardwareMemoryFits?Best quantNeedsMax tok/s (est.)
RTX 3060 12GB12 GBFitsQ2_K9.8 GB508details
RTX 4060 Ti 16GB16 GBFitsQ3_K_M12.3 GB316details
RTX 309024 GBFitsQ5_K_M18.4 GB668details
RTX 409024 GBFitsQ5_K_M18.4 GB719details
RTX 509032 GBFitsQ8_026.9 GB857details
RTX 5070 Ti16 GBFitsQ3_K_M12.3 GB984details
RX 7900 XTX24 GBFitsQ5_K_M18.4 GB685details
Mac M4 Pro 48GB48 GBFitsQ8_026.9 GB131details
Mac M4 Max 128GB128 GBFitsF1649.7 GB139details
Mac M3 Ultra 512GB512 GBFitsF1649.7 GB208details
DGX Spark128 GBFitsF1649.7 GB69details
RTX 407012 GBFitsQ2_K9.8 GB711details
RTX 5060 Ti 16GB16 GBFitsQ3_K_M12.3 GB492details
RTX 40608 GBNo———details
Arc B58012 GBFitsQ2_K9.8 GB643details
RTX 3060 Ti8 GBNo———details
RTX 30708 GBNo———details
RTX 3080 10GB10 GBNo———details
RTX 3080 12GB12 GBFitsQ2_K9.8 GB1287details
RTX 3090 Ti24 GBFitsQ5_K_M18.4 GB719details
RTX 4060 Ti 8GB8 GBNo———details
RTX 4070 Super12 GBFitsQ2_K9.8 GB711details
RTX 4070 Ti12 GBFitsQ2_K9.8 GB711details
RTX 4070 Ti Super16 GBFitsQ3_K_M12.3 GB738details
RTX 408016 GBFitsQ3_K_M12.3 GB788details
RTX 4080 Super16 GBFitsQ3_K_M12.3 GB808details
RTX 50608 GBNo———details
RTX 507012 GBFitsQ2_K9.8 GB948details
RTX 508016 GBFitsQ3_K_M12.3 GB1055details
2x RTX 30902×24 GBFitsQ8_026.9 GB448details
2x RTX 40902×24 GBFitsQ8_026.9 GB482details
2x RTX 50902×32 GBFitsF1649.7 GB456details
RTX A600048 GBFitsQ8_026.9 GB367details
RTX 6000 Ada48 GBFitsQ8_026.9 GB459details
RTX PRO 6000 Blackwell96 GBFitsF1649.7 GB456details
L40S48 GBFitsQ8_026.9 GB413details
A100 40GB40 GBFitsQ8_026.9 GB744details
A100 80GB80 GBFitsF1649.7 GB519details
H100 PCIe80 GBFitsF1649.7 GB509details
H100 SXM80 GBFitsF1649.7 GB853details
H200141 GBFitsF1649.7 GB1222details
RX 76008 GBNo———details
RX 7700 XT12 GBFitsQ2_K9.8 GB609details
RX 7800 XT16 GBFitsQ3_K_M12.3 GB685details
RX 7900 GRE16 GBFitsQ3_K_M12.3 GB633details
RX 7900 XT20 GBFitsQ4_K_M15.8 GB667details
RX 907016 GBFitsQ3_K_M12.3 GB703details
RX 9070 XT16 GBFitsQ3_K_M12.3 GB703details
Ryzen AI Max+ 395 128GB128 GBFitsF1649.7 GB65details
Arc A770 16GB16 GBFitsQ3_K_M12.3 GB615details
Mac M1 16GB16 GBTightQ2_K9.8 GB96details
Mac M2 24GB24 GBFitsQ3_K_M12.3 GB110details
Mac M3 24GB24 GBFitsQ3_K_M12.3 GB110details
Mac M4 16GB16 GBTightQ2_K9.8 GB169details
Mac M4 32GB32 GBFitsQ5_K_M18.4 GB86details
Mac M1 Pro 32GB32 GBFitsQ5_K_M18.4 GB143details
Mac M3 Pro 36GB36 GBFitsQ6_K21.0 GB93details
Mac M4 Pro 24GB24 GBFitsQ3_K_M12.3 GB300details
Mac M4 Pro 64GB64 GBFitsQ8_026.9 GB131details
Mac M1 Max 64GB64 GBFitsQ8_026.9 GB191details
Mac M2 Max 96GB96 GBFitsF1649.7 GB102details
Mac M3 Max 128GB128 GBFitsF1649.7 GB102details
Mac M4 Max 64GB64 GBFitsQ8_026.9 GB261details
Mac M2 Ultra 192GB192 GBFitsF1649.7 GB204details
Mac M3 Ultra 256GB256 GBFitsF1649.7 GB208details

Questions

How much VRAM does Trinity Mini need?

About 15.8 GB at 4-bit (Q4_K_M) with an 8K context: 14.8 GB of weights (real GGUF file size), 0.2 GB of KV cache and 0.8 GB of runtime overhead. At 8-bit it needs about 26.9 GB.

What is the smallest GPU that runs Trinity Mini?

At Q4_K_M and 8K context, the smallest single consumer GPU in our table that fits it is the RX 7900 XT (20 GB).

Can I run Trinity Mini on a Mac?

Yes. At Q4_K_M with 8K context it fits on a Mac M4 32GB or larger, using the default macOS GPU memory limit.

Model facts

Hugging Face repoarcee-ai/Trinity-Mini
Parameters26,123,974,400 (26B) · ~2.0B active (computed from config)
Layers / hidden size32 / 2048
AttentionGQA + sliding window; 32 heads, 4 KV heads, head dim 128; sliding window 2048
KV cache per token (FP16)64.0 KiB (sliding layers capped at the window)
Max context131,072 tokens
Licenseother
Released on HFDec 1, 2025
Downloads (30d) / likes11,429 / 207

Sources: Hugging Face model API · config.json (arcee-ai/Trinity-Mini) · GGUF files · GPU specs from vendor pages (list). Data fetched Nov 17, 2026. Spot an error? Report it.