Qwen3.8 27B Coder390 EfficientThink Opus5.5 GPT6Astra Grok4.7 DSV4Pro K3 SFT RLOO MTP DFlash2 GPU requirements: VRAM and cheapest GPU
Qwen3.8 27B Coder390 EfficientThink Opus5.5 GPT6Astra Grok4.7 DSV4Pro K3 SFT RLOO MTP DFlash2 has 27.3B parameters, published in GGUF. It reads images as well as text; each image adds tokens to the context. It is released under apache-2.0, which allows commercial use.
- Parameters
- 27.3B
- Smallest footprint (INT4)
- about 19 GB
- Cheapest GPU per hour (FP16)
- $0.50
- Type
- Vision
nerkyor/Qwen3.8-27B-Coder390-EfficientThink-Opus5.5-GPT6Astra-Grok4.7-DSV4Pro-K3-SFT-RLOO-MTP-DFlash2
- Parameters 27.3B
- Type Vision
- Task image-text-to-text
- Licence apache-2.0
- Downloads 876.9K
- Likes 587
Calculating for Qwen3.8 27B Coder390 EfficientThink Opus5.5 GPT6Astra Grok4.7 DSV4Pro K3 SFT RLOO MTP DFlash2
More settings
Qwen3.8 27B Coder390 EfficientThink Opus5.5 GPT6Astra Grok4.7 DSV4Pro K3 SFT RLOO MTP DFlash2 needs about 63 GB of GPU memory
Inference (vLLM), FP16 / BF16, 8K tokens context, one request at a time. Some of the model settings were missing, so the KV cache is an estimate.
Where the memory goes
- Weights 55 GB
- KV cache 1.7 GB
- Runtime overhead 6.6 GB
- Total 63 GB
GPU counts keep 10% of each card free for spikes. Plan extra if you expect long prompts or many users at once.
| Rank | Configuration | Per GPU / hr | GPUs needed | Total / hr | Total / month (24/7) | Provider |
|---|---|---|---|---|---|---|
| 1 | 1 × RTX PRO 6000 96GBCheapestBlackwell · GDDR7 | $0.50 | 1 | $0.50 | $365 | RunPod (opens in a new tab)1x RTX PRO 6000 MaxQ |
| 2 | 3 × RTX 3090 24GBAmpere · GDDR6X | $0.24 | 3sold as 4 GPUs; you need 3 | $0.96 | $702 | Vast.ai (opens in a new tab)4x RTX 3090 (median of 9 offers) |
| 3 | 1 × A100 80GBAmpere · HBM2e | $1.12 | 1 | $1.12 | $820 | Vast.ai (opens in a new tab)1x A100 SXM4 (median of 16 offers) |
| 4 | 2 × A40 48GBAmpere · GDDR6 | $0.59 | 2 | $1.18 | $861 | RunPod (opens in a new tab)2x A40 |
| 5 | 2 × A100 40GBAmpere · HBM2 | $0.61 | 2 | $1.22 | $890 | Vast.ai (opens in a new tab)2x A100 SXM4 (median of 8 offers) |
| 6 | 2 × RTX A6000 48GBAmpere · GDDR6 | $0.69 | 2 | $1.38 | $1,007 | RunPod (opens in a new tab)2x RTX A6000 |
| 7 | 1 × H100 80GBHopper · HBM3 | $1.39 | 1 | $1.39 | $1,015 | Lium (opens in a new tab)1x H100 80GB HBM3 |
| 8 | 2 × L40S 48GBAda Lovelace · GDDR6 | $0.80 | 2 | $1.60 | $1,169 | Vast.ai (opens in a new tab)2x L40S (median of 6 offers) |
| 9 | 1 × GH200 96GBHopper · HBM3 | $1.99 | 1 | $1.99 | $1,453 | Vultr (opens in a new tab)vbm-72c-480gb-gh200-gpu (bare metal) |
| 10 | 2 × RTX 6000 Ada 48GBAda Lovelace · GDDR6 | $1.09 | 2 | $2.18 | $1,591 | RunPod (opens in a new tab)2x RTX 6000 Ada |
| 11 | 3 × RTX 4090 24GBAda Lovelace · GDDR6X | $0.58 | 3sold as 4 GPUs; you need 3 | $2.32 | $1,696 | Vast.ai (opens in a new tab)4x RTX 4090 (median of 12 offers) |
| 12 | 1 × MI300X 192GBCDNA 3 · HBM3 | $2.39 | 1 | $2.39 | $1,745 | RunPod (opens in a new tab)1x MI300X |
| 13 | 3 × L4 24GBAda Lovelace · GDDR6 | $0.69 | 3sold as 4 GPUs; you need 3 | $2.76 | $2,015 | RunPod (opens in a new tab)4x L4 |
| 14 | 1 × H200 141GBHopper · HBM3e | $2.90 | 1 | $2.90 | $2,117 | Lium (opens in a new tab)1x H200 NVL |
| 15 | 1 × H100 NVL 94GBHopper · HBM3 | $3.19 | 1 | $3.19 | $2,329 | RunPod (opens in a new tab)1x H100 NVL |
| 16 | 3 × RTX 5090 32GBBlackwell · GDDR7 | $0.80 | 3sold as 4 GPUs; you need 3 | $3.20 | $2,338 | Vast.ai (opens in a new tab)4x RTX 5090 (median of 35 offers) |
| 17 | 1 × B200 180GBBlackwell · HBM3e | $5.60 | 1 | $5.60 | $4,088 | Lium (opens in a new tab)1x B200 |
| 18 | 3 × A10 24GBAmpere · GDDR6 | $1.42 | 3sold as 4 GPUs; you need 3 | $5.67 | $4,141 | AWS (opens in a new tab)g5.12xlarge |
| 19 | 5 × V100 16GBVolta · HBM2 | $0.79 | 5sold as 8 GPUs; you need 5 | $6.32 | $4,614 | Lambda (opens in a new tab)8x Tesla V100 |
| 20 | 5 × T4 16GBTuring · GDDR6 | $0.98 | 5sold as 8 GPUs; you need 5 | $7.82 | $5,712 | AWS (opens in a new tab)g4dn.metal |
| 21 | 1 × B300 288GBBlackwell Ultra · HBM3e | $8.00 | 1 | $8.00 | $5,840 | Lium (opens in a new tab)1x B300 SXM6 |
| 22 | 1 × GB200 186GBBlackwell · HBM3e | $10.50 | 1sold as 4 GPUs; you need 1 | $42.00 | $30,660 | CoreWeave (opens in a new tab)GB200 NVL72 |
| Provider | Instance | Per GPU / hr | GPUs needed | Total / hr | Total / month (24/7) |
|---|---|---|---|---|---|
| RunPod (opens in a new tab)Cheapest | 1x RTX PRO 6000 MaxQus | $0.50 | 1 × RTX PRO 6000 96GB | about $0.50 | about $365 |
| Vast.ai (opens in a new tab) | 4x RTX 3090 (median of 9 offers)worldwide marketplace | $0.24 | 3 × RTX 3090 24GB sold as 4 GPUs; you need 3 | about $0.96 | about $702 |
| Lium (opens in a new tab) | 1x RTX PRO 6000 Blackwell Server Editionworldwide marketplace | $1.19 | 1 × RTX PRO 6000 96GB | about $1.19 | about $869 |
| Nebius (opens in a new tab) | 1x RTX PRO 6000 96GBeu/us | $1.80 | 1 × RTX PRO 6000 96GB | about $1.80 | about $1,314 |
| Vultr (opens in a new tab) | vbm-72c-480gb-gh200-gpu (bare metal)us | $1.99 | 1 × GH200 96GB | about $1.99 | about $1,453 |
| Lambda (opens in a new tab) | 2x A6000us | $1.09 | 2 × RTX A6000 48GB | about $2.18 | about $1,591 |
| AWS (opens in a new tab) | g7e.2xlargeus-east-1 | $3.36 | 1 × RTX PRO 6000 96GB | about $3.36 | about $2,455 |
| Azure (opens in a new tab) | NC24ads_A100_v4eastus | $3.67 | 1 × A100 80GB | about $3.67 | about $2,681 |
| Google Cloud (opens in a new tab) | g2-standard-48us-central1 | $1.00 | 3 × L4 24GB sold as 4 GPUs; you need 3 | about $4.00 | about $2,921 |
| CoreWeave (opens in a new tab) | GH200us | $6.50 | 1 × GH200 96GB | about $6.50 | about $4,745 |
Prices collected Nov 16, 2026. Every cost is approximate: on-demand, US dollars, mostly US regions (marketplaces such as Vast.ai and Lium are worldwide).
Memory by precision and workload
| Precision | Inference | LoRA fine-tune | QLoRA fine-tune | Full fine-tune |
|---|---|---|---|---|
| FP16 / BF16 | 63 GB | 79 GB | – | 490 GB |
| INT8 | 34 GB | – | – | – |
| INT4 | 19 GB | – | 35 GB | – |
Cheapest GPU by precision
Inference with an 8K context and one request at a time.
FP16 / BF16
about 63 GB · 8K tokens
The weights as published. Nothing lost, but the largest footprint.
1 × RTX PRO 6000 96GB
RunPod (opens in a new tab) · 1x RTX PRO 6000 MaxQ
- Per GPU / hr
- $0.50
- GPUs needed
- 1
- Total / hr
- about $0.50
- per month, 24/7, approx.
- $365
INT8
about 34 GB · 8K tokens
Weights at half the size. Most chat and RAG work shows no visible difference.
1 × RTX A6000 48GB
Lium (opens in a new tab) · 1x RTX A6000
- Per GPU / hr
- $0.42
- GPUs needed
- 1
- Total / hr
- about $0.42
- per month, 24/7, approx.
- $307
INT4
about 19 GB · 8K tokens
Weights at a quarter of the size. Small drops show up first in maths and code.
1 × RTX 3090 24GB
Lium (opens in a new tab) · 1x RTX 3090 Ti
- Per GPU / hr
- $0.16
- GPUs needed
- 1
- Total / hr
- about $0.16
- per month, 24/7, approx.
- $117
Inference vs fine-tuning
Inference (FP16)
1×
of the memory inference needs
63 GB
1 × RTX PRO 6000 96GB · about $0.50
LoRA fine-tune
1.3×
of the memory inference needs
79 GB
1 × RTX PRO 6000 96GB · about $0.50
QLoRA fine-tune
0.5×
of the memory inference needs
35 GB
1 × RTX A6000 48GB · about $0.42
Full fine-tune
7.8×
of the memory inference needs
490 GB
3 × MI300X 192GB · about $9.56
Serving it at FP16 fits on 1 × RTX PRO 6000 96GB (RunPod). LoRA needs 1 × RTX PRO 6000 96GB (RunPod), QLoRA 1 × RTX A6000 48GB (Lium) and a full fine-tune 3 × MI300X 192GB (RunPod). Training keeps activations and optimiser state in memory as well as the weights, which is why it so often needs more GPUs than serving the same model.
Will it run on a Mac or a single RTX 4090?
RTX 4090 (24 GB)
Yes, at INT4 with an 8K context: about 19 GB fits in the card’s 24 GB.
Mac with Apple silicon
Yes, with enough memory. At INT4 it needs about 19 GB, so a Mac with 32 GB or more of unified memory (macOS lets the GPU use about three quarters of it). Expect slower generation than on a data-centre GPU.
Similar models
Open models of a similar size.
- ClefVision27.4B · about 62 GB FP16 · 79K ↓GPU needs
- Qwen3.8-27BVision27.8B · about 63 GB FP16 · 6.5M ↓GPU needs
- Qwen3.6Vision27.8B · about 63 GB FP16 · 2.4M ↓GPU needs
- Qwen3.5 27BVision27.8B · about 63 GB FP16 · 1.6M ↓GPU needs
- AutoTrust Decider Pro Vision 27BVision27.8B · about 63 GB FP16 · 893.9K ↓GPU needs
- DiffusionGemma 26B A4BVision25.8B · about 58 GB FP16 · 581K ↓GPU needs
Qwen3.8 27B Coder390 EfficientThink Opus5.5 GPT6Astra Grok4.7 DSV4Pro K3 SFT RLOO MTP DFlash2: common questions
How much VRAM does Qwen3.8 27B Coder390 EfficientThink Opus5.5 GPT6Astra Grok4.7 DSV4Pro K3 SFT RLOO MTP DFlash2 need?
About 63 GB at FP16, 34 GB at INT8 and 19 GB at INT4, with an 8K context and one request at a time. Longer contexts and more parallel requests add KV cache on top.
What is the cheapest GPU to run Qwen3.8 27B Coder390 EfficientThink Opus5.5 GPT6Astra Grok4.7 DSV4Pro K3 SFT RLOO MTP DFlash2?
1 × RTX PRO 6000 96GB (about $0.50 per GPU / hour) on RunPod, at about $0.50 an hour for FP16 / BF16 inference with an 8K context. Prices collected Nov 16, 2026; they change often.
Can I fine-tune Qwen3.8 27B Coder390 EfficientThink Opus5.5 GPT6Astra Grok4.7 DSV4Pro K3 SFT RLOO MTP DFlash2 on a single GPU?
Yes, with QLoRA: it needs about 35 GB, which fits one 80 GB GPU. LoRA needs about 79 GB and a full fine-tune about 490 GB.
How are these numbers calculated?
Weights are parameters times bytes per parameter (2 at FP16, 1 at INT8, 0.5 at INT4). The KV cache comes from the model’s layers, KV heads, head size and context length. We add about 10% plus 1 GB for the runtime and fill each GPU to 90%. Prices are on-demand list prices in US regions, collected from each provider’s public pricing.
Can Qwen3.8 27B Coder390 EfficientThink Opus5.5 GPT6Astra Grok4.7 DSV4Pro K3 SFT RLOO MTP DFlash2 run on a consumer GPU like the RTX 4090?
Yes, at INT4 with an 8K context: about 19 GB fits in the card’s 24 GB.
Does Qwen3.8 27B Coder390 EfficientThink Opus5.5 GPT6Astra Grok4.7 DSV4Pro K3 SFT RLOO MTP DFlash2 fit on a single H100 80GB?
Yes, at FP16 / BF16, INT8, INT4 with an 8K context.
How many GPUs do I need for Qwen3.8 27B Coder390 EfficientThink Opus5.5 GPT6Astra Grok4.7 DSV4Pro K3 SFT RLOO MTP DFlash2?
At FP16 with an 8K context: 1 × 80 GB, 2 × 48 GB or 3 × 24 GB GPUs. At INT4 you need about a quarter of the memory for the weights.
How much VRAM does Qwen3.8 27B Coder390 EfficientThink Opus5.5 GPT6Astra Grok4.7 DSV4Pro K3 SFT RLOO MTP DFlash2 need at INT8?
About 34 GB: 29 GB for the weights, 1.7 GB for an 8K KV cache and 4 GB of runtime overhead.
How much does it cost to run Qwen3.8 27B Coder390 EfficientThink Opus5.5 GPT6Astra Grok4.7 DSV4Pro K3 SFT RLOO MTP DFlash2 24/7 for a month?
About $365 on the cheapest setup we found (1 × RTX PRO 6000 96GB (about $0.50 per GPU / hour), $0.50 an hour for 730 hours). The cheapest on AWS, Google Cloud or Azure is about $2,455 a month.
FP16 vs INT8 vs INT4: which should I use?
FP16 keeps full quality and needs the most memory (about 63 GB here). INT8 halves the weights with little loss for most chat work. INT4 (AWQ, GPTQ or GGUF Q4) cuts them to a quarter, about 19 GB in total, with a small drop in quality that shows most in maths and code.
Can Qwen3.8 27B Coder390 EfficientThink Opus5.5 GPT6Astra Grok4.7 DSV4Pro K3 SFT RLOO MTP DFlash2 run on a Mac?
Yes, with enough memory. At INT4 it needs about 19 GB, so a Mac with 32 GB or more of unified memory (macOS lets the GPU use about three quarters of it). Expect slower generation than on a data-centre GPU.
vLLM or llama.cpp?
vLLM is built for serving many requests at once on NVIDIA and AMD GPUs and reserves most of the GPU memory for KV cache. llama.cpp runs GGUF files on almost anything, including Macs and CPUs, and suits one user or a small team. Memory needs are similar for the same precision.
Prices collected Nov 16, 2026. Every cost is approximate: on-demand, US dollars, mostly US regions (marketplaces such as Vast.ai and Lium are worldwide).
How we calculate
- Weights. Parameters times bytes per parameter: 2 at FP16, 1 at INT8, 0.5 at INT4, plus about 5% for the quantisation scales at INT8 and INT4.
- KV cache. 2 × layers × KV heads × head size × context × requests × 2 bytes, from the model’s config.json. Where a setting is missing we estimate it and say so.
- Overhead. About 10% plus 1 GB for the runtime, in line with what vLLM reserves.
- Fine-tuning. LoRA keeps 16-bit weights and adds a small adapter, its optimiser and activations. QLoRA does the same over 4-bit weights. A full fine-tune needs about 16 bytes per parameter with mixed-precision Adam (weights, gradients and optimiser states), plus activations.
- GPUs. We fill each GPU to 90% of its memory and stop at 8 per node. Costs are on-demand list prices in US regions; marketplace clouds show the median offer.