Ant Group, China · Released October 22, 2026New
Ling-3.0-flash-Sante has 127.5B parameters (5.1B active per token). In BF16 its weights alone take 232 GiB; quantized to 4-bit, about 70.0 GiB. Each 32K-token request adds 0.25 GiB of KV cache. Cheapest way to run it today: 6 × RTX A4000 (16 GB) at $0.528 per hour. It calls tools and has a thinking mode you can switch off.
Size your own setup in the calculator →Weights plus the KV cache of one request (BF16 cache) and the model's fixed recurrent state, in GiB. Add 1 to 3 GiB for the inference engine itself.
| Weights format | Weights only | + 8K context | + 32K context | + 128K context | + 256K context |
|---|---|---|---|---|---|
| BF16 / FP16 | 232 | 232 | 232 | 233 | 234 |
| FP8 / INT8 | 117 | 117 | 117 | 118 | 119 |
| 4-bit (GGUF Q4_K_M) | 70.0 | 70.2 | 70.4 | 71.1 | 72.1 |
Exact size of each quantization in AtomicChat/Ling-3.0-flash-Sante-GGUF (opens in a new tab), and the smallest GPU that runs it with llama.cpp, Ollama or LM Studio and an 8K-token context.
| Quantization | File size | Smallest GPU |
|---|---|---|
| Q8_0 | 127 GiB | More than 96 GB: see the calculator |
| Q6_K | 103 GiB | More than 96 GB: see the calculator |
| IQ4_XS | 72.8 GiB | RTX PRO 6000 Blackwell (96 GB) |
| IQ3_S | 56.9 GiB | RTX PRO 6000 Blackwell (96 GB) |
| IQ2_M | 48.8 GiB | RTX PRO 6000 Blackwell (96 GB) |
| IQ1_S | 33.2 GiB | RTX 6000 Ada (48 GB) |
Read every day from Hugging Face. Split files are added together; vision projectors are left out.
Fewest GPUs that fit the weights plus one 8K-token request. Data-center GPUs use vLLM (90% of memory usable), the others llama.cpp. Several GPUs split the model with tensor or pipeline parallelism.
| GPU | BF16 / FP16 | 8-bit | 4-bit |
|---|---|---|---|
| GeForce RTX 4090 24 GB | 16 (several machines) | 6 | 4 |
| GeForce RTX 5090 32 GB | 8 | 4 | 3 |
| RTX PRO 6000 Blackwell 96 GB | 3 | 2 | 1 |
| Mac M5 Max 128 GB 96 GB usable | 3 | 2 | 1 |
| Ryzen AI Max+ 395 128 GB 96 GB usable | 3 | 2 | 1 |
| DGX Spark (GB10) 128 GB | 2 | 1 | 1 |
| NVIDIA H100 80 GB | 4 | 2 | 2 |
| NVIDIA H200 141 GB | 2 | 1 | 1 |
| NVIDIA B200 (HGX) 180 GB | 2 | 1 | 1 |
8-bit: FP8 with vLLM (INT8 on GPUs without FP8), Q8_0 with llama.cpp. 4-bit: AWQ / GPTQ with vLLM, Q4_K_M with llama.cpp, both counted at the Q4_K_M size.
For each GPU: the fewest cards that fit Ling-3.0-flash-Sante with one 8K-token request, the most faithful weight format at that count, and the cheapest on-demand price (2026-10-23).
| Setup | Per hour | Weights | Provider | Per month | Rent |
|---|---|---|---|---|---|
| 6 × RTX A4000 16 GB | $0.528 | Q4_K_M | Vast.ai | $385.44 | Rent 6 × RTX A4000 on Vast.ai (opens in a new tab) → |
| 4 × GeForce RTX 3090 24 GB | $0.564 | Q4_K_M | Vast.ai | $411.72 | Rent 4 × GeForce RTX 3090 on Vast.ai (opens in a new tab) → |
| 2 × RTX A6000 48 GB | $0.668 | Q4_K_M | Vast.ai | $487.64 | Rent 2 × RTX A6000 on Vast.ai (opens in a new tab) → |
| 4 × RTX A5000 24 GB | $0.856 | Q4_K_M | Vast.ai | $624.88 | Rent 4 × RTX A5000 on Vast.ai (opens in a new tab) → |
| 2 × NVIDIA L40S 48 GB | $0.934 | 4-bit | Vast.ai | $681.82 | Rent 2 × NVIDIA L40S on Vast.ai (opens in a new tab) → |
| 4 × RTX A4500 20 GB | $1.00 | Q4_K_M | RunPod | $730.00 | Rent 4 × RTX A4500 on RunPod (opens in a new tab) → |
| 2 × NVIDIA A40 48 GB | $1.07 | 4-bit | Vast.ai | $779.64 | Rent 2 × NVIDIA A40 on Vast.ai (opens in a new tab) → |
| 4 × RTX PRO 4000 Blackwell 24 GB | $1.08 | Q4_K_M | Vast.ai | $788.40 | Rent 4 × RTX PRO 4000 Blackwell on Vast.ai (opens in a new tab) → |
On-demand prices from AWS, Google Cloud, Azure, RunPod, Vast.ai, Oracle Cloud, OVHcloud, Scaleway, Verda, Akash and UpCloud, checked every hour. Serving many users needs more KV cache, so more memory: size it in the calculator. Some links are affiliate links: StudioTV may earn a commission, at no extra cost to you.
AWS, Google Cloud and UpCloud prices: Spare Cores (opens in a new tab), CC BY-SA 4.0 (opens in a new tab), shared under the same license.
Reasoning. Thinking mode you can switch on or off. With vLLM or SGLang, set it per request: "chat_template_kwargs": {"enable_thinking": false} turns it off, true on. Reasoning tokens take context and KV cache like any other token: leave room for them.
Architecture: 42 layers · MLA (512 + 64 latent) on 7 · 35 Kimi Delta Attention. Its KV cache is 6.0× smaller than if every layer used full attention, because only 7 of its 42 layers keep a KV cache; the others hold a small fixed-size state. Only 5.1B parameters are active per token, but all 127B must sit in memory: VRAM depends on the total, speed on the active part.
| Context per request | BF16 cache | FP8 cache |
|---|---|---|
| 8K tokens | 0.06 GiB | 0.03 GiB |
| 32K tokens | 0.25 GiB | 0.12 GiB |
| 128K tokens | 0.98 GiB | 0.49 GiB |
| 256K tokens | 1.97 GiB | 0.98 GiB |
Plus 72.5 MiB of recurrent state per request, whatever the context.
In BF16 the weights alone take 232 GiB (249 GB). In FP8 that is 117 GiB, and about 70.0 GiB with 4-bit quantization (Q4_K_M). Each request then adds KV cache: 0.06 GiB at 8K tokens and 0.25 GiB at 32K (BF16 cache).
No. Even in 4-bit it needs 4 RTX 4090s. In 4-bit it fits on a single RTX PRO 6000 Blackwell (96 GB).
No. Even in 4-bit it needs 3 RTX 5090s.
Yes. In 4-bit (Q4_K_M with llama.cpp) it needs a Mac with 96 GB of unified memory or more, of which macOS gives the GPU about 72 GB, for one request with an 8K-token context. In 8-bit (Q8_0) it needs 192 GB or more.
Yes. On a DGX Spark (128 GB) it fits in Q8_0 (117 GiB with an 8K-token context); on a Ryzen AI Max+ 395 with 128 GB, whose GPU gets up to 96 GB, it fits in Q4_K_M (70.2 GiB with an 8K-token context), with llama.cpp and one request.
On 2026-10-23, the cheapest on-demand setup is 6 × RTX A4000 (16 GB) with Q4_K_M weights on Vast.ai, at $0.528 per hour (about $385.44 per month). Next: 4 × GeForce RTX 3090 (24 GB) with Q4_K_M weights on Vast.ai, at $0.564 per hour. Sized for one 8K-token request; prices are checked every hour.
With one request and an 8K-token context: 4 in BF16, 2 in FP8 and 2 in 4-bit (vLLM, 90% of memory usable). Serving many users at once needs more memory for their KV caches: the calculator sizes that for you.
Yes, Ling-3.0-flash-Sante supports tool (function) calling: it can ask your application to run a function and use the result.
Thinking mode you can switch on or off. With vLLM or SGLang, set it per request with "chat_template_kwargs": {"enable_thinking": false} (or true). Reasoning tokens count toward the context and the KV cache.
For a model card or a README: the 4-bit size and the smallest GPU it fits on, linked to this page. Data as JSON: /api/models/ling-3-0-flash-sante.json.
Ant Group, China · Released October 22, 2026[](https://studiotvai.com/vram-requirements/ling-3-0-flash-sante)Same family (hybrid linear attention): Qwen3.5 4B · MiniCPM-V-4.7 1B · Raon-1 mini preview · Qwen3.5 9B · Qwen3.8 27B · Qwen3.6 35B-A3B · Qwen3.5 122B-A10B · Qwen3.8-Flash-Next · LFM2.5 8B-A1B · LFM2 24B-A2B · Ornith 1.5 35B-A3B · Ornith 1.5 9B · Clef · Clef Flash · D1 3B · JEV 27B VL · LightOnOCR-3 4B · Agnes-3.0-Qwen · Agnes-3.0-Flash · Qwen3.6 27B · Qwen3.5 27B · Qwen3.5 35B-A3B · MiniCPM-V-4.7 35B-A3B · LFM2.5 24B-A2B · Ornith-1.5 122B · Sage-A1.5 · MLExpert · Agents-A1 4B
Llama 3.1 8B · Llama 3 8B · Llama 3.2 3B · Nemotron 3.5 Parse · Gemma 2 9B · Gemma 3 4B · LLaVA 1.6 Mistral 7B · Mistral Small 3.2 24B · Qwen3 32B · Llama 3.3 70B · Llama 3.1 405B · Hy-Memory 30B-A3B · Xing4.0 29B-A4B · Qwen3.5 4B · MiniCPM-V-4.7 1B · Raon-1 mini preview · Qwen3.5 9B · Qwen3.8 27B · Qwen3.6 35B-A3B · Qwen3.5 122B-A10B · Qwen3.8-Flash-Next · LFM2.5 8B-A1B · LFM2 24B-A2B · Gemma 4 12B · Gemma 4 31B · Gemma 4 26B-A4B · gpt-oss 20B · gpt-oss 120B · Llama 4 Scout 17B-16E · Llama 4 Maverick 17B-128E · Kolibri-1 · Recursive Flint · Nemotron Labs Dory 1.0 4B · Step 3.5 Flash · Muse Glimmer 30B · Nemotron 3 Nano 30B-A3B · Nemotron 3 Super 120B-A12B · Nemotron 3.5 Super VL 120B-A12B · Nemotron 3.5 Bolt 74B-A9B · MiniMax M3.1 Flash · MiniMax M3 · MiniMax M2.7 · DeepSeek R1 · Kimi K2.6 · GLM-5.3 · GLM-5.3-Flash · DeepSeek V4 Flash · DeepSeek V4 Pro · DeepSeek V4.1 Flash · MiMo-V2.6 Pro · MiMo-V2.6 Flash · Muse Spark Open · Beam 1.0 501B-A23B · Kimi K3 · Mistral Large 4 · Qwen3-Coder-Next 80B-A3B · Qwen3-Coder 30B-A3B · Qwen3-Coder 480B-A35B · DeepSeek V4 Flash Vision Exp · Qwen3.8 2.4T-A95B · Ornith 1.5 397B · Ornith 1.5 35B-A3B · Ornith 1.5 9B · Atria Dawn Preview · Intern S2 397B · Clef · Clef Flash · D1 3B · Humanizer · JEV 27B VL · Mellum2.1 12B-A2.5B · LightOnOCR-3 4B · Spark-X2.5 4B · Agnes-3.0-Qwen · Youtu-Parsing-Omni · Agnes-3.0-Flash · Mistral 7B v0.3 · Mistral Small 3.1 24B · Mistral Nemo · Ministral 8B · DeepSeek V3 · DeepSeek R1 Qwen3 8B · DeepSeek R1 Distill Qwen 32B · DeepSeek V3.1 · DeepSeek R1 Distill Qwen 14B · DeepSeek R1 Distill Qwen 7B · DeepSeek R1 Distill Llama 8B · Qwen3 4B · Qwen3 8B · Qwen2.5 7B · Qwen3 14B · Qwen2.5 3B · Qwen3.6 27B · Qwen3.5 27B · Qwen2.5-Coder 7B · Qwen2.5 32B · Qwen2.5-Coder 14B · Qwen3.5 35B-A3B · Qwen3 30B-A3B · Qwen2.5 14B · Qwen2.5-Coder 32B · Kimi K2.7 Code · Kimi K2.5 · Kimi K2 · GLM-5.2 · GLM-5.1 · GLM-5 · MiniMax M2.5 · MiniMax M2 · MiMo-V2.5 · Ornith-1.0 397B · Qwen2.5-Coder 3B · Qwen3 235B-A22B · Qwen3.5 397B-A17B · Gemma 4 E4B · Gemma 4 E2B · Nemotron 3 Nano 4B · Nemotron 3.5 Lightning 30B-A3B · MiniCPM-V-4.7 35B-A3B · LFM2.5 24B-A2B · GLM-4.7 · Qwen3-VL 8B · Qwen2.5-VL 7B · Qwen3-VL 4B · Qwen2.5-VL 3B · Qwen2.5-VL 32B · Qwen2-VL 7B · GLM-4.5-Air · GLM-4.5 · phi-4 · Phi-3.5-mini · Phi-4-mini · Phi-3-mini-128k · Unlimited-OCR · Ling-3.1-flash · Step 5 Preview · Hy3 · Ornith-1.5 122B · Spark-X2.6 4B · Sage-A1.5 · MLExpert · Agents-A1 4B · Atria-Dawn
Estimates, not guarantees: computed from the official config.json with the same engine as the LLM VRAM Calculator. Real usage depends on your engine version and settings.