Capacity
57 GiBfree
Fits
Room to grow — 128K context would still fit at this concurrency.
What runs on your hardware, and how comfortably.
Estimates from a roofline model calibrated against published measurements. Treat them as a band, not a promise. Model catalog generated .
4.7M downloads/mo
Most-downloaded first, by Hugging Face downloads over the 30 days before the catalog was generated on 2026-08-21 — a snapshot, not a live count.
US launch list price: $3,999 for the represented machine. Before tax. Checked 2026-08-16.
Fully coherent CPU/GPU pool, so no VRAM split to model. Strong prefill against modest bandwidth — the clearest case for reporting decode and TTFT separately.
How gpt-oss ships. Experts at 4-bit, attention untouched.
Dequantizes a quantized checkpoint to FP16 before the matmul, so a card’s low-precision peak is out of reach.
The window each user gets: prompt plus everything generated so far. Multiplied by the number of users, this is what sizes the KV cache.
How many sequences are in flight at once. Each one holds its own cache, so this multiplies memory directly.
How much of that context is already filled when generation starts. Part of the context, not extra — it sets how long you wait for the first token.
How many of this machine to shard the model across. llama.cpp / Ollama runs whole layers on each device in turn, so this buys capacity, not speed — one device’s bandwidth is the ceiling however many you add.
63 GiB / 120 GiB
| Component | Size | Share of ceiling | What it is |
|---|---|---|---|
| Weights | 61 GiB | 51% | Fixed. Set by parameter count and quantization — the one part that does not move when you change usage. |
| KV cache | 1.1 GiB | 1% | Grows with context x concurrency. The term that turns a comfortable fit into an OOM. |
| Overhead | 0.6 GiB | 1% | Runtime context, kernels and activation workspace. Small, but it is why 100% of nominal is never available. |
| Allocatable ceiling | 120 GiB | 100% | What the runtime can actually hand the model. |
57 GiBfree
Fits
Room to grow — 128K context would still fit at this concurrency.
36tok/s per user
Fast
Bound by weight bandwidth. Lower quantization or faster memory is what moves this.
5.7 s
Noticeable
1438 tok/s prompt processing, bound by compute on the linear layers.
6 of 7 workloads
Every model in the catalog, at every format and runtime DGX Spark (GB10) can use — ranked.
One person, back and forth, reading as it types.
Ranked by verdict first, then by parameter count, then by how little the weights are compressed, then by decode rate. Parameter count is a proxy for capability and not a measurement of it — Headroom knows what runs, not what is good.
The format that grades best, and the widest of those — narrowing only as far as the verdict requires. Only formats the runtime can actually load.
Swept 105 model and runtime pairings on DGX Spark (GB10) with a 16-bit cache at one user, each graded at the prompt its own workload sends. Change the hardware, the cache or the user count above and the list moves.
The flags for this exact placement — the layer split, the cache precision and the window above, in each launcher’s own spelling.
llama-server \
-m <path to your gpt-oss 120B MXFP4 (expert-only) .gguf> \
-c 32768 \
-ngl 37 \
-ctk f16 -ctv f16llama-server serves; it does not measure. The llama-bench command below is the measurement form of this same placement.
Flags checked against upstream documentation on 2026-08-01.
(
set -e
(set -C; cat > headroom-gpt-oss-120b-mxfp4-expert-only.Modelfile) <<'EOF'
FROM <path to your gpt-oss 120B MXFP4 (expert-only) .gguf>
PARAMETER num_ctx 32768
EOF
ollama create headroom-gpt-oss-120b-mxfp4-expert-only -f headroom-gpt-oss-120b-mxfp4-expert-only.Modelfile && ollama run headroom-gpt-oss-120b-mxfp4-expert-only
)Ollama ships no benchmark client. It runs llama.cpp, so the llama-bench command in this panel measures the same engine on the same GGUF — that is the form to submit.
Flags checked against upstream documentation on 2026-08-01.
llama-bench measures; it does not serve. The llama-server command above is the serving form of this same placement.
llama-bench \
-m <path to your gpt-oss 120B MXFP4 (expert-only) .gguf> \
-p 8192 \
-n 0 \
-ngl 37 \
-ctk f16 -ctv f16 \
-o md
llama-bench \
-m <path to your gpt-oss 120B MXFP4 (expert-only) .gguf> \
-p 0 \
-n 128 \
-d 32640 \
-ngl 37 \
-ctk f16 -ctv f16 \
-o mdFlags checked against upstream documentation on 2026-08-01.
Every figure on this page is a model, calibrated on two measured machines and asserted within ±30%. Paste your own llama-bench output and see both numbers side by side.
Parsed here in your browser. The text never leaves this page, and nothing is sent anywhere unless you choose to open the pre-filled issue below.
1 of 56 comfortable