Moonshot AI · Chat and reasoning · 49.1B parameters · 3B active · mit
Kimi Linear 48B A3B Instruct
Needs 59 GB at FP8 / INT8. The smallest real configuration today is 2× Tesla V100 32GB at Vast.ai, $0.14 per hour, about $99 a month at 730 hours.
Parameters49.1B3B active per token
Context—maximum published
Checkpoint98 GBBF16, as published
Licensemitcheck before commercial use
Released—documented date
Smallest configuration · FP8 / INT8
2× Tesla V100 32GB64 GB VRAM · 92% used
- Per hour
- $0.14
- Per month
- $99
- Cheapest at
Vast.ai
Memory · 3 precisionsHow much VRAM it needs
Weights plus runtime overhead. The context dimension appears once the model's architecture is on file.
| Precision | Weights | Weights + headroom |
|---|---|---|
| BF16published checkpoint | 98 GB | 118 GB |
| FP8 / INT8estimated | 49 GB | 59 GB |
| INT4 / MXFP4estimated | 29 GB | 36 GB |
Cross-checked on Hugging Face Oct 28, 2026: 49.1B parameters in the safetensors headers, 98.3 GB of weight files, attention layout from config.json (27 layers, latent rank 512). 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 = 1152 bytes per token per layer × 27 full-attention layers (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 |
|---|---|---|---|---|---|
| Tesla V100 32GBBest fit | 2× 32 GB64 GB total | 92% tight | $0.14 | $99 | Vast.aiTrust 78 · read Nov 2 · receipt · also Cyfuture AI $1.13, io.net $2.52 |
| Tesla P100 | 4× 16 GB64 GB total | 92% tight | $0.19 | $142 | Vast.aiTrust 78 · read Nov 2 · receipt · also Edgevana $0.43 |
| RTX 3090 | 4× 24 GB96 GB total | 61% | $0.25 | $183 |  Clore.aiTrust 49 · read Nov 2 · receipt · also Vast.ai $0.56, Edgevana $0.92 |
| RTX 5060 Ti | 4× 16 GB64 GB total | 92% tight | $0.35 | $258 | Clore.aiTrust 49 · read Nov 2 · receipt · also Vast.ai $0.40, Edgevana $0.68 |
| RTX A4000 | 4× 16 GB64 GB total | 92% tight | $0.36 | $266 | Vast.aiTrust 78 · read Nov 2 · receipt · also AtmosCompute $0.66, Spheron $3.84 |
| RTX A6000 | 2× 48 GB96 GB total | 61% | $0.38 | $274 | Clore.aiTrust 49 · read Nov 2 · receipt · also Lium $0.84, Spheron $0.89 |
| Tesla V100 16GB | 4× 16 GB64 GB total | 92% tight | $0.38 | $275 | Vast.aiTrust 78 · read Nov 2 · receipt · also CloudRift $1.00, AtmosCompute $1.30 |
| Quadro RTX 5000 | 4× 16 GB64 GB total | 92% tight | $0.40 | $294 | Vast.aiTrust 78 · read Nov 2 · receipt |
| RTX 3060 | 8× 12 GB96 GB total | 61% | $0.43 | $312 | Vast.aiTrust 78 · read Nov 2 · receipt · also Edgevana $0.64, AtmosCompute $1.01 |
| A800 80GB PCIe | 1× 80 GB80 GB total | 74% | $0.47 | $342 | Vast.aiTrust 78 · read Nov 2 · receipt |
| Quadro RTX 8000 | 2× 48 GB96 GB total | 61% | $0.48 | $354 | Vast.aiTrust 78 · read Nov 2 · receipt · also Edgevana $0.56 |
| RTX 3070 | 8× 8 GB64 GB total | 92% tight | $0.50 | $364 | Clore.aiTrust 49 · read Nov 2 · receipt · also Vast.ai $5.23 |
| L40 | 2× 48 GB96 GB total | 61% | $0.54 | $391 | Vast.aiTrust 78 · read Nov 2 · receipt · also io.net $1.32, GPU.ai $1.56 |
| RTX 2080 Ti | 8× 11 GB88 GB total | 67% | $0.59 | $432 | Vast.aiTrust 78 · read Nov 2 · receipt · also Edgevana $0.68 |
| T4 | 4× 16 GB64 GB total | 92% tight | $0.59 | $434 | Vast.aiTrust 78 · read Nov 2 · receipt · also iwinv $0.78, NHN Cloud $2.46 |
| A100 80GB | 1× 80 GB80 GB total | 74% | $0.60 | $440 | Vast.aiTrust 78 · read Nov 2 · receipt · also packet.ai $0.89, GPUBank $0.95 |
| RTX 5060 | 8× 8 GB64 GB total | 92% tight | $0.65 | $472 | Clore.aiTrust 49 · read Nov 2 · receipt |
| RTX PRO 6000 | 1× 96 GB96 GB total | 61% | $0.66 | $482 |  packet.ainot scored · read Nov 2 · receipt · also Runpod $0.69, HyperAI $0.80 |
| Titan RTX | 4× 24 GB96 GB total | 61% | $0.70 | $508 | Vast.aiTrust 78 · read Nov 2 · receipt |
| RTX 5070 Ti | 4× 16 GB64 GB total | 92% tight | $0.70 | $508 | Vast.aiTrust 78 · read Nov 2 · receipt · also Clore.ai $0.80, Edgevana $0.80 |
| RTX 4070 Ti Super | 4× 16 GB64 GB total | 92% tight | $0.75 | $548 | Clore.aiTrust 49 · read Nov 2 · receipt · also AtmosCompute $1.54 |
| L40S | 2× 48 GB96 GB total | 61% | $0.76 | $555 |  LiumTrust 40 · read Nov 2 · receipt · also Vast.ai $1.07, CloudRift $1.26 |
| A100 40GB | 2× 40 GB80 GB total | 74% | $0.77 | $566 | Vast.aiTrust 78 · read Nov 2 · receipt · also iwinv $1.83, AtmosCompute $1.93 |
| RTX 5090 | 2× 32 GB64 GB total | 92% tight | $0.80 | $586 | Vast.aiTrust 78 · read Nov 2 · receipt · also Clore.ai $0.96, Edgevana $1.35 |
| The 24 cheapest of 70 GPU types that fit. Every GPU, any precision and context, in the finder | |||||
Throughput estimates · Oct 4, 2026Tokens 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 an estimate at the prices read on Oct 4, 2026; it is not recomputed here.
| GPU | Cards | Precision | tok/s per user | per 1M output tokens · estimate |
|---|---|---|---|---|
| MI300X | 1 | FP8 | 190 | $0.157 |
| RTX 4090 | 4 | FP8 | 70.3 | $0.169 |
| RTX 5090 | 2 | FP8 | 102 | $0.175 |
| H100 SXM | 1 | FP8 | 200 | $0.207 |
| H200 | 1 | FP8 | 225 | $0.210 |
| B200 | 1 | FP8 | 295 | $0.219 |
| RTX PRO 6000 | 1 | FP8 | 151 | $0.228 |
| H100 PCIe | 1 | FP8 | 154 | $0.239 |
| L40S | 2 | FP8 | 74.1 | $0.300 |
| A100 80GB | 2 | BF16 | 101 | $0.397 |
cards, precision and speed estimated from memory bandwidth and model size; cost recomputed by the site from verified prices · cheapest per output at the prices of Oct 4, 2026: MI300XWhy estimates, and their limits →
Kimi Linear 48B A3B Instruct · related