# SmolLM2 135M Instruct on an Apple M3 Ultra (512GB)

Yes — the unquantized weights fit with room for 8k tokens of context.

Canonical page: https://makerportal.ai/lab/llm-vram/smollm2-135m-instruct/apple-m3-ultra-512gb
Page title: SmolLM2 135M Instruct VRAM on M3 Ultra 512GB — fits at BF16

This markdown document and the HTML page above are rendered from the same solved values at build time, by the same functions. Nothing here is written by a language model and nothing is fetched at request time.

## Key figures

- **Verdict:** Fits at FP16 — 384.0 GiB usable of 512 GB
- **Usable memory:** 384.0 GiB — 75% of the card's 512 GB — an assumption for unified memory, not a single fraction applied to every device
- **Weights at FP16:** 257 MiB — 135M parameters
- **Weights (Q4_K_M):** 101 MiB — 257 MiB at FP16 · 135M params · 6 of 6 quant rows are measured files
- **KV cache:** 22.5 KiB per token at FP16 — 30 layers × 3 KV heads × 64 dimensions
- **Speed ceiling:** 1789 tok/s — 819 GB/s ÷ bytes read per token
- **Largest context that fits:** 8,192 tokens at FP16 / BF16
- **Trained context:** 8,192 — the cap the KV-cache figures are held to
- **Card bandwidth:** 819 GB/s — 512 GB LPDDR5 unified

## Every quant, against this card

Every quantization of SmolLM2 135M Instruct against the 384.0 GiB usable on an Apple M3 Ultra (512GB). "Measured" rows are the byte size of a real published file; "nominal" rows are the parameter count times the published bits-per-weight.

| Precision | Bits/weight | Weights | Source | Fits | Max context | Ceiling |
|---|---|---|---|---|---|---|
| FP16 / BF16 | 16.00 | 257 MiB | measured file | yes | 8,192 (model cap) | 1789 tok/s |
| Q8_0 | 8.61 | 138 MiB | measured file | yes | 8,192 (model cap) | 2455 tok/s |
| Q6_K | 8.23 | 132 MiB | measured file | yes | 8,192 (model cap) | 2504 tok/s |
| Q5_K_M | 6.67 | 107 MiB | measured file | yes | 8,192 (model cap) | 2722 tok/s |
| Q4_K_M | 6.27 | 101 MiB | measured file | yes | 8,192 (model cap) | 2784 tok/s |
| Q3_K_M | 5.56 | 89 MiB | measured file | yes | 8,192 (model cap) | 2902 tok/s |

## Questions this page answers

### How much VRAM does SmolLM2 135M Instruct need?

257 MiB for the weights at FP16 — 135M parameters at two bytes each — and 101 MiB at Q4_K_M. The KV cache is on top of that and is not a fixed number: this model spends 22.5 KiB per token of context, so 8,192 tokens costs a further 180 MiB. An Apple M3 Ultra (512GB) makes 384.0 GiB of its 512 GB available on the assumption below.

### Can an Apple M3 Ultra (512GB) run SmolLM2 135M Instruct?

Yes — the unquantized weights fit with room for 8k tokens of context. That is the weights and the KV cache together, against 384.0 GiB of usable memory.

### How fast will SmolLM2 135M Instruct run on an Apple M3 Ultra (512GB)?

No faster than 1789 tokens/second at FP16 / BF16, and in practice below it. Decoding is memory-bound: every token reads all 257 MiB of weights plus the cache, and this card moves 819 GB/s. That division is the ceiling — no kernel, runtime or driver beats it, and a real runtime typically reaches 60–80% of it.

### Why does the context length change how much memory SmolLM2 135M Instruct needs?

Because the KV cache holds one key and one value vector per token, per layer, for the whole conversation, and it is allocated separately from the weights. This model has 30 layers and 3 key/value heads of 64 dimensions, shared across 9 query heads — grouped-query attention, which divides the cache by 3. That works out at 22.5 KiB per token. Parameter count tells you nothing about this number.

## Method and limits

Weight bytes are the byte size of the real published file wherever one exists, and the model's exact parameter count times the published llama.cpp bits-per-weight where it does not. That distinction is on every row above and it matters at both ends: a sub-1B model's Q4_K_M file runs a third larger than the nominal figure because k-quants keep its embedding tables at higher precision, and an already-4-bit release cannot be quantized upward at all. The KV cache is 2 · layers · kv_heads · head_dim · bytes per token, summed over layers with each sliding-window layer capped at its window. The speed figure is a roofline bound, not a benchmark: bandwidth divided by bytes read per token, which no runtime exceeds and every runtime falls short of. The usable fraction of card memory is an assumption: 75% here, which is what this lane assumes for unified memory, where the OS and window server share the same pool — the other class assumes 92%, so it is not one number applied to every device. That is the only assumed input on this page; every other figure is computed from the model config and the card's published specification. Nothing on this page is written by a language model.

## Related tool

[LLM VRAM & KV-Cache Footprint Calculator](https://makerportal.ai/lab/llm-vram-kvcache-calculator) — Interactive VRAM planner — change the model geometry, quantization, context length and card capacity and read the weight bytes, KV-cache bytes and the bandwidth ceiling on decode speed.

---

Source: MakerPortal — https://makerportal.ai/lab/llm-vram/smollm2-135m-instruct/apple-m3-ultra-512gb. Free to quote and cite with attribution and a link to the canonical page.