Allen Institute for AI · allen-ai/olmo-3-32b-think
| Provider | Allen Institute for AI |
|---|---|
| Family | olmo |
| Type | llm-reasoning |
| Status | active |
| Released | 2025-11-19 |
| Updated | 2026-06-25 |
| Parameters | 32.2B |
| Open weights | yes |
| Relationship | Model |
|---|---|
| Is a finetune of | allenai/Olmo-3-32B-Think-DPO |
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 | 32.23 GB | ~456047.0 | ~912093.9 q4 |
| NVIDIA GB200 Grace Blackwell Superchip | 372.0 GB | 16000.0 GB/s | bf16 | 64.47 GB | ~173.7 | ~694.9 q4 |
| AMD Instinct MI355X | 288.0 GB | 8000.0 GB/s | bf16 | 64.47 GB | ~86.9 | ~347.5 q4 |
| NVIDIA B200 | 180.0 GB | 7700.0 GB/s | bf16 | 64.47 GB | ~83.6 | ~334.4 q4 |
| Google TPU7x (Ironwood) | 192.0 GB | 7380.0 GB/s | bf16 | 64.47 GB | ~80.1 | ~320.5 q4 |
| AMD Instinct MI325X | 256.0 GB | 6000.0 GB/s | bf16 | 64.47 GB | ~65.1 | ~260.6 q4 |
| AMD Instinct MI300X | 192.0 GB | 5300.0 GB/s | bf16 | 64.47 GB | ~57.5 | ~230.2 q4 |
| NVIDIA H200 SXM | 141.0 GB | 4800.0 GB/s | bf16 | 64.47 GB | ~52.1 | ~208.5 q4 |
| NVIDIA H100 NVL | 94.0 GB | 3900.0 GB/s | bf16 | 64.47 GB | ~42.3 | ~169.4 q4 |
| NVIDIA H100 SXM | 80.0 GB | 3350.0 GB/s | fp8 | 32.23 GB | ~72.8 | ~145.5 q4 |
| AMD Instinct MI250X | 128.0 GB | 3200.0 GB/s | bf16 | 64.47 GB | ~34.7 | ~139.0 q4 |
| Google TPU v5p | 95.0 GB | 2765.0 GB/s | bf16 | 64.47 GB | ~30.0 | ~120.1 q4 |
| NVIDIA A100 80GB SXM | 80.0 GB | 2039.0 GB/s | fp8 | 32.23 GB | ~44.3 | ~88.6 q4 |
| NVIDIA H100 PCIe | 80.0 GB | 2000.0 GB/s | fp8 | 32.23 GB | ~43.4 | ~86.9 q4 |
| NVIDIA GeForce RTX 5090 | 32.0 GB | 1792.0 GB/s | q5 | 20.15 GB | ~62.3 | ~77.8 q4 |
| Google TPU v6e (Trillium) | 32.0 GB | 1638.0 GB/s | q5 | 20.15 GB | ~56.9 | ~71.1 q4 |
| AMD Instinct MI210 | 64.0 GB | 1600.0 GB/s | fp8 | 32.23 GB | ~34.7 | ~69.5 q4 |
| NVIDIA A100 40GB SXM | 40.0 GB | 1555.0 GB/s | q6 | 24.18 GB | ~45.0 | ~67.5 q4 |
| Google TPU v4 | 32.0 GB | 1200.0 GB/s | q5 | 20.15 GB | ~41.7 | ~52.1 q4 |
| NVIDIA GeForce RTX 4090 | 24.0 GB | 1008.0 GB/s | q4 | 16.12 GB | ~43.8 | ~43.8 q4 |
| AMD Radeon RX 7900 XTX | 24.0 GB | 960.0 GB/s | q4 | 16.12 GB | ~41.7 | ~41.7 q4 |
| NVIDIA GeForce RTX 3090 | 24.0 GB | 936.0 GB/s | q4 | 16.12 GB | ~40.7 | ~40.7 q4 |
| NVIDIA L40S | 48.0 GB | 864.0 GB/s | fp8 | 32.23 GB | ~18.8 | ~37.5 q4 |
| Apple M3 Ultra | 512.0 GB | 819.0 GB/s | bf16 | 64.47 GB | ~8.9 | ~35.6 q4 |
| Apple M2 Ultra | 192.0 GB | 800.0 GB/s | bf16 | 64.47 GB | ~8.7 | ~34.7 q4 |
| Apple M4 Max | 128.0 GB | 546.0 GB/s | bf16 | 64.47 GB | ~5.9 | ~23.7 q4 |
| Apple M1 Max | 64.0 GB | 400.0 GB/s | fp8 | 32.23 GB | ~8.7 | ~17.4 q4 |
| Apple M2 Max | 96.0 GB | 400.0 GB/s | bf16 | 64.47 GB | ~4.3 | ~17.4 q4 |
| Apple M3 Max | 128.0 GB | 400.0 GB/s | bf16 | 64.47 GB | ~4.3 | ~17.4 q4 |
| NVIDIA L4 | 24.0 GB | 300.0 GB/s | q4 | 16.12 GB | ~13.0 | ~13.0 q4 |
| Apple M4 Pro | 64.0 GB | 273.0 GB/s | fp8 | 32.23 GB | ~5.9 | ~11.9 q4 |
| Apple M4 | 32.0 GB | 120.0 GB/s | q5 | 20.15 GB | ~4.2 | ~5.2 q4 |
Taken from the card's availability section.
| Platform | Id on that platform |
|---|---|
| Hugging Face | allenai/Olmo-3-32B-Think |
| Benchmark | Evaluated as | Score | Evidence date | Source kind | |
|---|---|---|---|---|---|
| aa_lcr | Olmo 3 32B Think | 0.0% | 2026-09-10 evaluated | independent evaluator | source |
| critpt | Olmo 3 32B Think | 0.0% | 2026-09-10 evaluated | independent evaluator | source |
| gpqa_diamond | Olmo 3 32B Think | 61.01% | 2026-09-10 evaluated | independent evaluator | source |
This card reports no benchmark scores yet.
| Benchmark | Catalogue standing | Score |
|---|