Agnes AI, Singapore · Released September 11, 2026New
Agnes-3.0-Flash has 33.1B parameters. In BF16 its weights alone take 60.8 GiB; quantized to 4-bit, about 18.4 GiB. Each 32K-token request adds 2.25 GiB of KV cache. Cheapest way to run it today: 1 × GeForce RTX 3090 (24 GB) at $0.136 per hour.
Size your own setup in the calculator →llama.cpp runs its GGUF build quimmedes/Agnes-3.0-Flash-XYZ-GGUF (opens in a new tab)
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 | 60.8 | 61.6 | 63.3 | 70.0 | 79.0 |
| FP8 / INT8 | 32.8 | 33.5 | 35.2 | 42.0 | 51.0 |
| 4-bit (GGUF Q4_K_M) | 18.4 | 19.1 | 20.8 | 27.5 | 36.5 |
Exact size of each quantization in 0xKitkat/Agnes-3.0-Flash-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 | 31.9 GiB | RTX 6000 Ada (48 GB) |
| Q6_K | 24.6 GiB | RTX 5090 (32 GB) |
| Q5_K_M | 21.4 GiB | RTX 4090 (24 GB) |
| Q4_K_M | 18.4 GiB | RTX 4090 (24 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 | FP8 / INT8 | 4-bit (GGUF Q4_K_M) |
|---|---|---|---|
| GeForce RTX 4090 24 GB | 3 | 2 | 1 |
| GeForce RTX 5090 32 GB | 3 | 2 | 1 |
| RTX PRO 6000 Blackwell 96 GB | 1 | 1 | 1 |
| Mac M5 Max 128 GB 96 GB usable | 1 | 1 | 1 |
| NVIDIA H100 80 GB | 1 | 1 | 1 |
| NVIDIA H200 141 GB | 1 | 1 | 1 |
| NVIDIA B200 (HGX) 180 GB | 1 | 1 | 1 |
For each GPU: the fewest cards that fit Agnes-3.0-Flash with one 8K-token request, the most faithful weight format at that count, and the cheapest on-demand price (2026-10-10).
| Setup | Per hour | Weights | Provider | Per month | Rent |
|---|---|---|---|---|---|
| 1 × GeForce RTX 3090 24 GB | $0.136 | 4-bit (GGUF Q4_K_M) | Vast.ai | $99.28 | Rent 1 × GeForce RTX 3090 on Vast.ai (opens in a new tab) → |
| 1 × GeForce RTX 4090 24 GB | $0.349 | 4-bit (GGUF Q4_K_M) | Vast.ai | $254.77 | Rent 1 × GeForce RTX 4090 on Vast.ai (opens in a new tab) → |
| 1 × NVIDIA A100 40GB | $0.428 | 4-bit (GGUF Q4_K_M) | Vast.ai | $312.44 | Rent 1 × NVIDIA A100 40GB on Vast.ai (opens in a new tab) → |
| 1 × GeForce RTX 5090 32 GB | $0.445 | 4-bit (GGUF Q4_K_M) | Vast.ai | $324.85 | Rent 1 × GeForce RTX 5090 on Vast.ai (opens in a new tab) → |
| 2 × GeForce RTX 5080 16 GB | $0.446 | 4-bit (GGUF Q4_K_M) | Vast.ai | $325.58 | Rent 2 × GeForce RTX 5080 on Vast.ai (opens in a new tab) → |
| 1 × RTX 6000 Ada 48 GB | $0.494 | FP8 / INT8 | Vast.ai | $360.62 | Rent 1 × RTX 6000 Ada on Vast.ai (opens in a new tab) → |
| 1 × NVIDIA L40S 48 GB | $0.536 | FP8 / INT8 | Vast.ai | $391.28 | Rent 1 × NVIDIA L40S on Vast.ai (opens in a new tab) → |
| 2 × NVIDIA L4 24 GB | $0.536 | FP8 / INT8 | Vast.ai | $391.28 | Rent 2 × NVIDIA L4 on Vast.ai (opens in a new tab) → |
On-demand prices from RunPod, Vast.ai, Verda and Azure, 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.
Architecture: 72 layers · 18 full attention (4 KV × 256) · 54 linear attention. Its KV cache is 4.0× smaller than if every layer used full attention, because only 18 of its 72 layers keep a KV cache; the others hold a small fixed-size state.
| Context per request | BF16 cache | FP8 cache |
|---|---|---|
| 8K tokens | 0.56 GiB | 0.28 GiB |
| 32K tokens | 2.25 GiB | 1.13 GiB |
| 128K tokens | 9.00 GiB | 4.50 GiB |
| 256K tokens | 18.0 GiB | 9.00 GiB |
Plus 165 MiB of recurrent state per request, whatever the context.
In BF16 the weights alone take 60.8 GiB (65 GB). In FP8 that is 32.8 GiB, and about 18.4 GiB with 4-bit quantization (Q4_K_M). Each request then adds KV cache: 0.56 GiB at 8K tokens and 2.25 GiB at 32K (BF16 cache).
Yes. In 4-bit (GGUF Q4_K_M) it fits on one RTX 4090 with an 8K-token context (19.1 GiB including the cache, with llama.cpp).
On 2026-10-10, the cheapest on-demand setup is 1 × GeForce RTX 3090 (24 GB) with 4-bit (GGUF Q4_K_M) weights on Vast.ai, at $0.136 per hour (about $99.28 per month). Next: 1 × GeForce RTX 4090 (24 GB) with 4-bit (GGUF Q4_K_M) weights on Vast.ai, at $0.349 per hour. Sized for one 8K-token request; prices are checked every hour.
With one request and an 8K-token context: 1 in BF16, 1 in FP8 and 1 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.
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/agnes-3-0-flash.json.
Agnes AI, Singapore · Released September 11, 2026[](https://studiotvai.com/vram-requirements/agnes-3-0-flash)Same family (hybrid linear attention): Qwen3.5 4B · 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
All models: Llama 3.1 8B · Mistral Small 3.2 24B · Qwen3 32B · Llama 3.3 70B · Llama 3.1 405B · Qwen3.5 4B · 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 · Nemotron 3 Nano 30B-A3B · Nemotron 3 Super 120B-A12B · 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 · 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
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.