Evidence: benchmark results vendor-reported
Trinity Mini — VRAM, speed & local setup
Written by Jakub Rusinowski · Last updated
Arcee AI's 26B-total, about 3B-active mixture-of-experts model with 128 routed experts (8 active) and one shared expert. It is tuned for reasoning, supports a 128K context, and Arcee publishes an official GGUF set that works with llama.cpp (b7061 or newer), plus vLLM, Transformers and LM Studio recipes. Licensed under OpenMDW-1.1. Benchmark charts on the card are vendor-reported.
Trinity Mini needs about 17 GB of VRAM at Q4_K_M — quantized weights plus framework overhead, before any KV cache. On Apple Silicon that figure comes out of unified memory.
VRAM and speed by quantization
Quoted against NVIDIA RTX 4090 (24 GB). Includes the KV cache at 8K context, so it reads higher than the headline figure.
| Quant | Memory | VRAM | Speed (est.) | Fit |
|---|---|---|---|---|
| Q2_K 2.63 bpw | 9.5 GB | ~275 tok/s | Fits | |
| Q3_K_M 3.41 bpw | 13 GB | ~248 tok/s | Fits | |
| Q4_K_M 4.83 bpw | 16.9 GB | ~211 tok/s | Fits | |
| Q5_K_M 5.67 bpw | 19.6 GB | ~194 tok/s | Fits | |
| Q6_K 6.56 bpw | 22.5 GB | ~179 tok/s | Tight | |
| Q8_0 8.50 bpw | 28.7 GB | ~23 tok/s | Offload | |
| F16 16.00 bpw | 53.2 GB | ~15 tok/s | Offload |
Black marker = usable memory on the NVIDIA RTX 4090 (24 GB). Estimates from the memory-bandwidth roofline on the methodology page. Trinity Mini VRAM calculator →
Get Trinity Mini running
The cheapest catalogued GPU that runs Trinity Mini is the AMD Radeon RX 7900 XT (20 GB).
or compare on Vast.ai from $0.35/hr (typical low · varies)
As an Amazon Associate we earn from qualifying purchases. Cloud GPU links are referral links — we may earn a commission at no extra cost to you.
How to run Trinity Mini
No verified Ollama library tag. Arcee publishes official GGUF files that work with llama.cpp (b7061 or newer) and LM Studio.
Specifications
Verified — Checked against the primary source — the model card or the vendor spec page — and corroborated by a second independent source.
- Parameters
- 26.1 Billion (3B active)
- Context window
- 131,072
- Architecture
- Mixture-of-Experts (128 routed + 1 shared, interleaved local and global attention)
- Provider
- Arcee AI
- Licence
- OpenMDW-1.1
- Specified at
- Q4_K_M
- System RAM
- 32 GB
- Record updated
- 2026-10-06
Commercial use permitted. No usage restrictions beyond attribution.
Limits and caveats
- Benchmark claims are vendor-reported and only available as a chart image.
- No verified Ollama library tag.
- Card recommends unusually low sampling temperature (0.15); other runtimes may need these settings explicitly.
Quality and use cases
Scores as published by the model’s authors or an independent evaluator — quality, not throughput, and not measured by us.
Other Trinity sizes
Trinity Mini — frequently asked questions
How much VRAM does Trinity Mini need?
About 17 GB at Q4_K_M — quantized weights plus framework overhead, before any KV cache. The cache grows with context length and is added on top; the table above folds it in. Apple Silicon counts unified memory toward the same figure.
Does Trinity Mini run on an RTX 4090 (24 GB)?
Yes. Trinity Mini needs about 17 GB at Q4_K_M, inside a 24 GB card, at an estimated 211 tokens/sec.
How do I run Trinity Mini locally?
No verified Ollama library tag. Arcee publishes official GGUF files that work with llama.cpp (b7061 or newer) and LM Studio. Running the published tag would send your prompts to a hosted GPU rather than your own machine.
What other sizes does Trinity come in?
Trinity Mini (17 GB), Trinity-Large-Thinking (241 GB). Every size shares the family's training and licence; the larger ones score higher and need proportionally more memory.