Model × hardware
Can Mac Studio M5 Max (32-core GPU) 36GB run OLMoE-1B-7B-0924-Instruct?
OLMoE-1B-7B-0924-Instruct at Q4_K_M fits the estimated Mac Studio M5 Max (32-core GPU) 36GB memory budget at 4,096 tokens: 5.42 GiB required versus 27.00 GiB usable. Estimated speed unavailable.
Selected formatQ4_K_M
Estimated requirement5.42 GiB
Usable device memory27.00 GiB
Estimated headroom21.58 GiB
Compare the recorded formats.
| Quant | Weight files | Estimated KV | Estimated total | Fit | Estimated generation |
|---|---|---|---|---|---|
| Q4_K_M | 3.92 GiB | 0.50 GiB | 5.42 GiB | fits | Estimated speed unavailable |
| Q5_K_M | 4.59 GiB | 0.50 GiB | 6.09 GiB | fits | Estimated speed unavailable |
| Q8_0 | 6.85 GiB | 0.50 GiB | 8.35 GiB | fits | Estimated speed unavailable |
| FP16 | 12.89 GiB | 0.50 GiB | 14.39 GiB | fits | Estimated speed unavailable |
What changes with a longer context?
| Context tokens | Estimated KV | Estimated total | Fit |
|---|---|---|---|
| 2,048 | 0.25 GiB | 5.17 GiB | fits |
| 4,096 | 0.50 GiB | 5.42 GiB | fits |
The formula for this estimate.
Weights GiB = exact recorded file bytes / 2^30. KV GiB = 2 × layers × KV heads × head dimension × context × KV bytes / 2^30. Required GiB = weights + KV + runtime overhead. Usable GiB = device GiB × usable fraction.
Assumptions and limitations
- Memory values labeled GB in the data contract are GiB (2^30 bytes); bandwidth uses decimal GB/s (10^9 bytes/s).
- The memory budget reserves 25.0% of device memory. Runtime overhead is an assumed 1 GiB.
- KV cache assumes 2 bytes per element, independently of weight quantization. Runtime cache layout and allocation can add memory.
- Memory fit does not establish runtime, driver, operating-system, architecture, or quantization support.
- All speed values are estimated single-stream decode proxies, not benchmarks. Prefill, batching, latency, and concurrent sessions are not predicted.
- Estimated speed sensitivity is not a confidence interval; actual speed can fall outside the range because of compute, cache traffic, kernels, thermal limits, and other workloads.
- Effective bytes per weight is the exact file-size / total-parameter proxy. Metadata, tensors with different precision, padding, and duplicated tensors make it different from nominal quantization bits.
- Weight memory includes every recorded MoE expert. Active parameters affect estimated decode traffic only, and uniform bytes per active parameter is an approximation.
- Active MoE parameter count is unknown, so estimated speed is unavailable.
- Estimated speed is withheld when memory fit or the parameters needed for the speed formula are unknown, or when the model exceeds the supported memory budget.
Source records
Official model metadata · Exact weight files · Maker hardware specifications
Model verified 2026-10-22; hardware verified 2026-10-22. This is an estimated memory and bandwidth calculation, not a hands-on benchmark.