OpenAI · Chat and reasoning · 117B parameters · 5.1B active · apache-2.0
gpt-oss-120b
Needs 79 GB at INT4 / MXFP4. The smallest real configuration today is 4× RTX 3090 at Clore.ai, $0.25 per hour, about $183 a month at 730 hours.
Parameters117B5.1B active per token
Context128k tokensmaximum published
Checkpoint65 GBINT4-packed, as published
Licenseapache-2.0check before commercial use
ReleasedAug 5, 2025documented date
Smallest configuration · INT4 / MXFP4
4× RTX 309096 GB VRAM · 82% used
- Per hour
- $0.25
- Per month
- $183
- Cheapest at
Clore.ai
Memory · 1 precision · 4 context lengthsHow much VRAM it needs
Weights plus runtime overhead, then the KV cache that grows with the context you serve.
| Precision | Weights | Weights + headroom | 4k context | 32k context | 128k context |
|---|---|---|---|---|---|
| INT4-packedpublished checkpoint | 65 GB | 79 GB | 79 GB | 80 GB | 84 GB |
| KV cacheone sequence, FP16 cache, included above | — | — | 1 GB | 2 GB | 5 GB |
Cross-checked on Hugging Face Oct 28, 2026: 116.8B parameters in the safetensors headers, 65.3 GB of weight files, attention layout from config.json (36 layers, 8 KV heads × 64, 18 sliding-window layers). Method: weights as published at the native precision, verified on Hugging Face 2026-10-28; other precisions estimated at 1 (8-bit) and 0.6 (4-bit) bytes per parameter; × 1.2 runtime headroom; KV cache = 2048 bytes per token per layer × 18 full-attention layers + 18 sliding-window layers capped at 128 tokens (FP16 cache, one sequence).config.json ↗
Smallest configuration per GPU · 110 GPUs · cheapest eligible on-demand offer per GPU count, every provider on fileWhat it costs to run, GPU by GPU
The fewest GPUs of one type whose total memory fits the model at short context, priced from the cheapest eligible on-demand offer for exactly that count (45-day window, newest reading per provider), with the next providers beside it. Monthly = 730 hours.
| GPU | Smallest config | VRAM used | $ / hour | $ / month | Cheapest at |
|---|---|---|---|---|---|
| RTX 3090Best fit | 4× 24 GB96 GB total | 82% | $0.25 | $183 | Clore.aiTrust 46 · read Nov 1 · receipt · also Vast.ai $0.58, Lium $0.88 |
| Tesla V100 32GB | 4× 32 GB128 GB total | 61% | $0.27 | $196 |  Vast.aiTrust 78 · read Nov 1 · receipt · also Clore.ai $0.41, Cyfuture AI $2.22 |
| RTX A6000 | 2× 48 GB96 GB total | 82% | $0.38 | $274 | Clore.aiTrust 46 · read Nov 1 · receipt · also Lium $0.84, Spheron $0.89 |
| Tesla P100 | 8× 16 GB128 GB total | 61% | $0.38 | $274 | Vast.aiTrust 78 · read Nov 1 · receipt |
| RTX 3060 | 8× 12 GB96 GB total | 82% | $0.43 | $312 | Vast.aiTrust 78 · read Nov 1 · receipt |
| A800 80GB PCIe | 1× 80 GB80 GB total | 98% tight | $0.47 | $342 | Vast.aiTrust 78 · read Nov 1 · receipt |
| Quadro RTX 8000 | 2× 48 GB96 GB total | 82% | $0.51 | $371 | Vast.aiTrust 78 · read Nov 1 · receipt |
| L40 | 2× 48 GB96 GB total | 82% | $0.54 | $391 | Vast.aiTrust 78 · read Nov 1 · receipt · also io.net $1.32, GPU.ai $1.56 |
| RTX 2080 Ti | 8× 11 GB88 GB total | 89% | $0.59 | $432 | Vast.aiTrust 78 · read Nov 1 · receipt · also Edgevana $0.68 |
| A100 80GB | 1× 80 GB80 GB total | 98% tight | $0.60 | $440 | Vast.aiTrust 78 · read Nov 1 · receipt · also packet.ai $0.89, GPUBank $0.95 |
| RTX PRO 6000 | 1× 96 GB96 GB total | 82% | $0.66 | $482 |  packet.ainot scored · read Nov 1 · receipt · also Runpod $0.69, HyperAI $0.80 |
| Titan RTX | 4× 24 GB96 GB total | 82% | $0.70 | $508 | Vast.aiTrust 78 · read Nov 1 · receipt |
| Tesla V100 16GB | 8× 16 GB128 GB total | 61% | $0.75 | $547 | Vast.aiTrust 78 · read Nov 1 · receipt · also CloudRift $2.32, Lambda $6.32 |
| RTX 5060 Ti | 8× 16 GB128 GB total | 61% | $0.75 | $549 | Vast.aiTrust 78 · read Nov 1 · receipt · also Clore.ai $0.82, Edgevana $1.51 |
| L40S | 2× 48 GB96 GB total | 82% | $0.76 | $555 |  LiumTrust 40 · read Nov 1 · receipt · also Vast.ai $1.28, VoltageGPU $1.52 |
| A100 40GB | 2× 40 GB80 GB total | 98% tight | $0.77 | $566 | Vast.aiTrust 78 · read Nov 1 · receipt · also Edgevana $1.29, AtmosCompute $1.45 |
| RTX A5000 | 4× 24 GB96 GB total | 82% | $0.80 | $586 | Vast.aiTrust 78 · read Nov 1 · receipt |
| RTX PRO 4000 | 4× 24 GB96 GB total | 82% | $0.83 | $608 | Clore.aiTrust 46 · read Nov 1 · receipt |
| RTX A4000 | 8× 16 GB128 GB total | 61% | $0.96 | $701 | LiumTrust 40 · read Nov 1 · receipt · also Vast.ai $0.96, VoltageGPU $1.92 |
| RTX PRO 5000 | 2× 48 GB96 GB total | 82% | $0.96 | $702 | Vast.aiTrust 78 · read Nov 1 · receipt · also Lium $1.70, AtmosCompute $2.04 |
| RTX 4070 Super | 8× 12 GB96 GB total | 82% | $0.97 | $708 | Vast.aiTrust 78 · read Nov 1 · receipt · also Edgevana $1.23, AtmosCompute $1.75 |
| RTX 6000 Ada | 2× 48 GB96 GB total | 82% | $0.99 | $722 | Vast.aiTrust 78 · read Nov 1 · receipt · also GPU.ai $1.30, io.net $1.37 |
| Gaudi 2 | 1× 96 GB96 GB total | 82% | $1.06 | $777 |  Cyfuture AInot scored · read Nov 1 · receipt |
| RTX 4000 Ada | 4× 20 GB80 GB total | 98% tight | $1.07 | $780 | Vast.aiTrust 78 · read Nov 1 · receipt · also iwinv $1.15, Akamai Cloud $2.96 |
| The 24 cheapest of 62 GPU types that fit. Every GPU, any precision and context, in the finder | |||||
Throughput estimates · internal study · 2026-10-04Tokens per second, and what a million costs
Chat and reasoning. Speed estimated from memory bandwidth and model size, never measured. The cost per 1M output tokens is the study's own estimate at the prices it read on 2026-10-04; it is not recomputed here.
| GPU | Cards | Precision | tok/s per user | per 1M output tokens · study estimate |
|---|---|---|---|---|
| MI300X | 1 | FP4 | 166 | $0.210 |
| RTX 4090 | 4 | FP4 | 56.9 | $0.237 |
| B200 | 1 | FP4 | 260 | $0.254 |
| A100 80GB | 1 | FP4 | 140 | $0.256 |
| RTX PRO 6000 | 1 | FP4 | 144 | $0.263 |
| H100 SXM | 1 | FP4 | 181 | $0.273 |
| RTX 5090 | 4 | FP4 | 68.1 | $0.281 |
| H200 | 1 | FP4 | 199 | $0.286 |
| H100 PCIe | 1 | FP4 | 145 | $0.308 |
| L40S | 2 | FP4 | 63.9 | $0.385 |
cards, precision and speed estimated from memory bandwidth and model size (internal study); cost recomputed by the site from verified prices · cheapest per output at study prices: MI300XWhy estimates, and their limits →
gpt-oss-120b · related