AI9Stars · ai9stars/g9v3-39a5b
| Provider | AI9Stars |
|---|---|
| Family | g9 |
| Type | llm-chat |
| Status | active |
| Released | 2026-07-21 |
| Updated | 2026-08-19 |
| Parameters | 39B |
| Open weights | yes |
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 | q6 | 29.23 GB | ~3822400.6 | ~5733600.9 q4 |
| NVIDIA GB200 Grace Blackwell Superchip | 372.0 GB | 16000.0 GB/s | bf16 | 77.93 GB | ~1092.1 | ~4368.5 q4 |
| AMD Instinct MI355X | 288.0 GB | 8000.0 GB/s | bf16 | 77.93 GB | ~546.1 | ~2184.2 q4 |
| NVIDIA B200 | 180.0 GB | 7700.0 GB/s | bf16 | 77.93 GB | ~525.6 | ~2102.3 q4 |
| Google TPU7x (Ironwood) | 192.0 GB | 7380.0 GB/s | bf16 | 77.93 GB | ~503.7 | ~2015.0 q4 |
| AMD Instinct MI325X | 256.0 GB | 6000.0 GB/s | bf16 | 77.93 GB | ~409.5 | ~1638.2 q4 |
| AMD Instinct MI300X | 192.0 GB | 5300.0 GB/s | bf16 | 77.93 GB | ~361.8 | ~1447.1 q4 |
| NVIDIA H200 SXM | 141.0 GB | 4800.0 GB/s | bf16 | 77.93 GB | ~327.6 | ~1310.5 q4 |
| NVIDIA H100 NVL | 94.0 GB | 3900.0 GB/s | fp8 | 38.97 GB | ~532.4 | ~1064.8 q4 |
| NVIDIA H100 SXM | 80.0 GB | 3350.0 GB/s | fp8 | 38.97 GB | ~457.3 | ~914.6 q4 |
| AMD Instinct MI250X | 128.0 GB | 3200.0 GB/s | bf16 | 77.93 GB | ~218.4 | ~873.7 q4 |
| Google TPU v5p | 95.0 GB | 2765.0 GB/s | fp8 | 38.97 GB | ~377.5 | ~754.9 q4 |
| NVIDIA A100 80GB SXM | 80.0 GB | 2039.0 GB/s | fp8 | 38.97 GB | ~278.4 | ~556.7 q4 |
| NVIDIA H100 PCIe | 80.0 GB | 2000.0 GB/s | fp8 | 38.97 GB | ~273.0 | ~546.1 q4 |
| NVIDIA GeForce RTX 5090 | 32.0 GB | 1792.0 GB/s | q4 | 19.48 GB | ~489.3 | ~489.3 q4 |
| Google TPU v6e (Trillium) | 32.0 GB | 1638.0 GB/s | q4 | 19.48 GB | ~447.2 | ~447.2 q4 |
| AMD Instinct MI210 | 64.0 GB | 1600.0 GB/s | fp8 | 38.97 GB | ~218.4 | ~436.8 q4 |
| NVIDIA A100 40GB SXM | 40.0 GB | 1555.0 GB/s | q6 | 29.23 GB | ~283.0 | ~424.6 q4 |
| Google TPU v4 | 32.0 GB | 1200.0 GB/s | q4 | 19.48 GB | ~327.6 | ~327.6 q4 |
| NVIDIA L40S | 48.0 GB | 864.0 GB/s | q6 | 29.23 GB | ~157.3 | ~235.9 q4 |
| Apple M3 Ultra | 512.0 GB | 819.0 GB/s | bf16 | 77.93 GB | ~55.9 | ~223.6 q4 |
| Apple M2 Ultra | 192.0 GB | 800.0 GB/s | bf16 | 77.93 GB | ~54.6 | ~218.4 q4 |
| Apple M4 Max | 128.0 GB | 546.0 GB/s | bf16 | 77.93 GB | ~37.3 | ~149.1 q4 |
| Apple M1 Max | 64.0 GB | 400.0 GB/s | fp8 | 38.97 GB | ~54.6 | ~109.2 q4 |
| Apple M2 Max | 96.0 GB | 400.0 GB/s | fp8 | 38.97 GB | ~54.6 | ~109.2 q4 |
| Apple M3 Max | 128.0 GB | 400.0 GB/s | bf16 | 77.93 GB | ~27.3 | ~109.2 q4 |
| Apple M4 Pro | 64.0 GB | 273.0 GB/s | fp8 | 38.97 GB | ~37.3 | ~74.5 q4 |
| Apple M4 | 32.0 GB | 120.0 GB/s | q4 | 19.48 GB | ~32.8 | ~32.8 q4 |
Taken from the card's availability section.
| Platform | Id on that platform |
|---|---|
| Hugging Face | ai9stars/G9v3-39A5B |
| Benchmark | Evaluated as | Score | Evidence date | Source kind | |
|---|---|---|---|---|---|
| aa_lcr | G9v3-39A5B | 65.33% | 2026-09-10 evaluated | independent evaluator | source |
| critpt | G9v3-39A5B | 0.29% | 2026-09-10 evaluated | independent evaluator | source |
| gdpval_aa | G9v3-39A5B | 34.44% | 2026-09-10 evaluated | independent evaluator | source |
| gpqa_diamond | G9v3-39A5B | 80.51% | 2026-09-10 evaluated | independent evaluator | source |
| scicode | G9v3-39A5B | 36.81% | 2026-09-10 evaluated | independent evaluator | source |
This card reports no benchmark scores yet.
| Benchmark | Catalogue standing | Score |
|---|