Prism ML · prism-ml/ternary-bonsai-27b
| Provider | Prism ML |
|---|---|
| Family | bonsai |
| Type | quantized-variant |
| Status | active |
| Released | 2026-07-14 |
| Updated | 2026-08-31 |
| Parameters | 27.4B |
| Open weights | yes |
| Relationship | Model |
|---|---|
| Is a quantized of | Qwen3.6 27B |
Chain Of Thought · tier-2 Function Calling · tier-2
Ordered by predicted speed. Bandwidth sets decode rate; memory decides whether it runs at all.
| Device | Memory | Bandwidth | Best quality | Weights | tok/s | Fastest |
|---|---|---|---|---|---|---|
| Cerebras WSE-3 | 44.0 GB | 21000000.0 GB/s | fp8 | 27.36 GB | ~537345.0 | ~1074689.9 q4 |
| NVIDIA GB200 Grace Blackwell Superchip | 372.0 GB | 16000.0 GB/s | bf16 | 54.71 GB | ~204.7 | ~818.8 q4 |
| AMD Instinct MI355X | 288.0 GB | 8000.0 GB/s | bf16 | 54.71 GB | ~102.4 | ~409.4 q4 |
| NVIDIA B200 | 180.0 GB | 7700.0 GB/s | bf16 | 54.71 GB | ~98.5 | ~394.1 q4 |
| Google TPU7x (Ironwood) | 192.0 GB | 7380.0 GB/s | bf16 | 54.71 GB | ~94.4 | ~377.7 q4 |
| AMD Instinct MI325X | 256.0 GB | 6000.0 GB/s | bf16 | 54.71 GB | ~76.8 | ~307.1 q4 |
| AMD Instinct MI300X | 192.0 GB | 5300.0 GB/s | bf16 | 54.71 GB | ~67.8 | ~271.2 q4 |
| NVIDIA H200 SXM | 141.0 GB | 4800.0 GB/s | bf16 | 54.71 GB | ~61.4 | ~245.6 q4 |
| NVIDIA H100 NVL | 94.0 GB | 3900.0 GB/s | bf16 | 54.71 GB | ~49.9 | ~199.6 q4 |
| NVIDIA H100 SXM | 80.0 GB | 3350.0 GB/s | bf16 | 54.71 GB | ~42.9 | ~171.4 q4 |
| AMD Instinct MI250X | 128.0 GB | 3200.0 GB/s | bf16 | 54.71 GB | ~40.9 | ~163.8 q4 |
| Google TPU v5p | 95.0 GB | 2765.0 GB/s | bf16 | 54.71 GB | ~35.4 | ~141.5 q4 |
| NVIDIA A100 80GB SXM | 80.0 GB | 2039.0 GB/s | bf16 | 54.71 GB | ~26.1 | ~104.3 q4 |
| NVIDIA H100 PCIe | 80.0 GB | 2000.0 GB/s | bf16 | 54.71 GB | ~25.6 | ~102.4 q4 |
| NVIDIA GeForce RTX 5090 | 32.0 GB | 1792.0 GB/s | q6 | 20.52 GB | ~61.1 | ~91.7 q4 |
| Google TPU v6e (Trillium) | 32.0 GB | 1638.0 GB/s | q6 | 20.52 GB | ~55.9 | ~83.8 q4 |
| AMD Instinct MI210 | 64.0 GB | 1600.0 GB/s | fp8 | 27.36 GB | ~40.9 | ~81.9 q4 |
| NVIDIA A100 40GB SXM | 40.0 GB | 1555.0 GB/s | fp8 | 27.36 GB | ~39.8 | ~79.6 q4 |
| Google TPU v4 | 32.0 GB | 1200.0 GB/s | q6 | 20.52 GB | ~40.9 | ~61.4 q4 |
| NVIDIA GeForce RTX 4090 | 24.0 GB | 1008.0 GB/s | q5 | 17.1 GB | ~41.3 | ~51.6 q4 |
| AMD Radeon RX 7900 XTX | 24.0 GB | 960.0 GB/s | q5 | 17.1 GB | ~39.3 | ~49.1 q4 |
| NVIDIA GeForce RTX 3090 | 24.0 GB | 936.0 GB/s | q5 | 17.1 GB | ~38.3 | ~47.9 q4 |
| NVIDIA L40S | 48.0 GB | 864.0 GB/s | fp8 | 27.36 GB | ~22.1 | ~44.2 q4 |
| Apple M3 Ultra | 512.0 GB | 819.0 GB/s | bf16 | 54.71 GB | ~10.5 | ~41.9 q4 |
| AMD Radeon RX 7900 XT | 20.0 GB | 800.0 GB/s | q4 | 13.68 GB | ~40.9 | ~40.9 q4 |
| Apple M2 Ultra | 192.0 GB | 800.0 GB/s | bf16 | 54.71 GB | ~10.2 | ~40.9 q4 |
| Apple M4 Max | 128.0 GB | 546.0 GB/s | bf16 | 54.71 GB | ~7.0 | ~27.9 q4 |
| Apple M1 Max | 64.0 GB | 400.0 GB/s | fp8 | 27.36 GB | ~10.2 | ~20.5 q4 |
| Apple M2 Max | 96.0 GB | 400.0 GB/s | bf16 | 54.71 GB | ~5.1 | ~20.5 q4 |
| Apple M3 Max | 128.0 GB | 400.0 GB/s | bf16 | 54.71 GB | ~5.1 | ~20.5 q4 |
| NVIDIA L4 | 24.0 GB | 300.0 GB/s | q5 | 17.1 GB | ~12.3 | ~15.4 q4 |
| Apple M4 Pro | 64.0 GB | 273.0 GB/s | fp8 | 27.36 GB | ~7.0 | ~14.0 q4 |
| Apple M4 | 32.0 GB | 120.0 GB/s | q6 | 20.52 GB | ~4.1 | ~6.1 q4 |
Taken from the card's availability section.
| Platform | Id on that platform |
|---|---|
| Hugging Face | prism-ml/Ternary-Bonsai-27B-gguf |
This card reports no benchmark scores yet.
| Benchmark | Catalogue standing | Score |
|---|