Tri-21B-Think
Files and memory at 8K context
| Quant | File | Files | + KV + runtime | Smallest fit | Source |
|---|---|---|---|---|---|
| Q4_K_M mykor/Tri-21B-Think-GGUF | 11.72 GiB | 1 | 14.3 GiB | 16 GB GPU | HF files |
| Q5_K_M mykor/Tri-21B-Think-GGUF | 13.72 GiB | 1 | 16.3 GiB | 20 GB GPU | HF files |
| Q8_0 mykor/Tri-21B-Think-GGUF | 20.52 GiB | 1 | 23.1 GiB | 24 GB GPU | HF files |
| FP16 KnutJaegersberg/Tri-21B-Think-gguf | 38.61 GiB | 1 | 41.2 GiB | 48 GB GPU | HF files |
No verified file: MLX 4-bit, MLX 8-bit. Sizes are exact file bytes (text weights; projectors listed separately). “+” = recurrent state or other terms unknown. Smallest fit adds 0.5 GiB spare; Macs use 75% of memory.
Quality
Artificial Analysis Intelligence Index
9.0 AA estimate
Variant: Tri-21B-Think · observed 2026-10-17 · source
Rank #65 of 75 researched 2026 models.
Arena text rating
Unknown
Scores describe the evaluated hosted variant, not each quantized file. AA and Arena are never averaged.
Reported speeds 0
No reported measurement yet. Submit one.
Community threads
- Device assertion errors with transformers Hugging Face · 2026-02-19
- Installation Video and Testing - Step by Step Hugging Face · 2026-03-13
Qualitative signals only; titles are not verified claims.