KAT Coder V2.5 Dev REAP 50 NVFP4A16 — Hardware Requirements & GPU Compatibility
ChatCodeFunctionsKAT Coder V2.5 Dev REAP 50 NVFP4A16 is a 18.5B-parameter open language model from Ttimms. It supports a context window of up to 262,144 tokens. At Q4_K_M it needs about 11.51 GB of VRAM — see which GPUs and Macs can run it below.
Runs great on a 24 GB GPU
Runs from 12 GB (Q4_K_M · 32K)Best on 24 GB (Q4_K_M · ~183 tok/s · 256K)Macs from 16 GB (Q4_K_M · ~43 tok/s)
Apache 2.0
Reference GPU · RTX 4090 24 GB
Estimates, ±20% · how we estimate
Run KAT Coder V2.5 Dev REAP 50 NVFP4A16
One recommended way. Every other method, OS and quant is in the step-by-step guide.Each way opens a step-by-step guide for your OS and quant.
Recommended
vllm serve Ttimms/KAT-Coder-V2.5-Dev-REAP-50-NVFP4A16 --max-model-len 16384vLLM repo with its own quantisation
Download: Weights on Hugging Face
Ways to runOther ways
each opens step-by-stepRun in cloud
Runs on RTX A4000 (16 GB) (4 GB headroom) · Q4_K_M
- Generation speed
- ~105 tok/s
- generation speed
- Cost per 1M output tokens
- $0.27
- per 1M output tokens
How Much VRAM Does KAT Coder V2.5 Dev REAP 50 NVFP4A16 Need?
Select a quantization level; the listed hardware adheres to it.
| Quantization | VRAM | +32K | On RTX 4090 (reference) | |
|---|---|---|---|---|
| 8.3 GB | 8.5 GB | 2-bit · K-quant — 2-bit quantization with K-quant improvements | Fits: 256K context · ~220 tok/s | |
| 9.4 GB | 9.7 GB | 3-bit medium — 3-bit medium quantization | Fits: 256K context · ~205 tok/s | |
| 11.5 GB | 11.8 GB | 4-bit medium — 4-bit medium quantization — most popular sweet spot | Fits: 256K context · ~183 tok/s | |
| 13.6 GB | 13.9 GB | 5-bit medium — 5-bit medium quantization — good quality/size tradeoff | Fits: 256K context · ~165 tok/s | |
| 15.7 GB | 15.9 GB | 6-bit, very good — 6-bit quantization, very good quality | Fits: 256K context · ~151 tok/s | |
| 18.9 GB | 19.2 GB | 8-bit, near-lossless — 8-bit quantization, near-lossless | Fits: 256K context · ~132 tok/s | |
| 37.5 GB | 37.7 GB | 16-bit · bfloat — Brain floating point 16 — preferred for training | Does not fit: needs 37.6 GB |
Selected Q4_K_M
4-bit medium quantization11.6GBat 16K context
- Weights
- 11.1 GB
- KV cache
- 0.2 GB
- Overhead
- 0.3 GB
Fits RTX 4090 (reference) · 12.4 GB free · ~183 tok/s
≥50 tok/s needs ≥185 GB/s · RTX 5090 ~248, RTX 3090 Ti ~183
Which Hardware Can Run KAT Coder V2.5 Dev REAP 50 NVFP4A16?
Q4_K_M · 11.5 GBKAT Coder V2.5 Dev REAP 50 NVFP4A16 (Q4_K_M) requires 11.5 GB of VRAM to load the model weights. For comfortable inference with headroom for KV cache and system overhead, 15+ GB is recommended. Using the full 262K context window can add up to 2.6 GB, bringing total usage to 14.1 GB. 46 GPUs can run it, including NVIDIA GeForce RTX 5090, NVIDIA GeForce RTX 3090 Ti, NVIDIA GeForce RTX 5080. 169 devices with unified memory can run KAT Coder V2.5 Dev REAP 50 NVFP4A16, including Mac Studio M5 Ultra (256 GB), Mac Studio M5 Ultra (512 GB), Mac Mini M2 Pro (16 GB).
12 GB
12 GB tier →fits · 0.4 GB free · up to 32K ctx
16 GPUs
RTX 3080 Ti~172Simulate output on RTX 3080 TiRTX 3080 12GB~172Simulate output on RTX 3080 12GBRTX 5070~141Simulate output on RTX 507020 GB
fits · 8.4 GB free · up to 256K ctx
1 GPU
RX 7900 XT~150Simulate output on RX 7900 XT32–40 GB
fits · 20.4 GB free · up to 256K ctx
1 GPU · 3 Mac classes · 9 device families
RTX 5090~248Simulate output on RTX 5090M5 Max*36 GB~68Simulate output on Mac Studio M5 Max (36 GB)M632 GB~39Simulate output on iMac 24" M6 (32 GB)RTX Spark N1X~76Simulate output on NVIDIA RTX Spark N1X Laptop (32 GB)Orin~55Simulate output on NVIDIA Jetson AGX Orin 32GB
By VRAM tier · typical GPUs and Macs at Q4_K_M
Similar models
1| Model | Score | VRAM | tok/s | Why it is similar |
|---|---|---|---|---|
| Internlm 20BInternLM | — | 45.3 GB | —needs > 24 GB | same size |
Specifications
- Publisher
- Ttimms
- Parameters
- 18.5B
- Active parameters
- 3.4B per token
- Architecture
- Qwen3_5MoeForConditionalGeneration
- Context Length
- 262,144 tokens
- Vocabulary Size
- 248,320
- Release Date
- 2026-08-20
- License
- Apache 2.0
Frequently Asked Questions
How much VRAM does KAT Coder V2.5 Dev REAP 50 NVFP4A16 need?
KAT Coder V2.5 Dev REAP 50 NVFP4A16 requires 11.5 GB of VRAM at Q4_K_M, or 37.5 GB at BF16. Full 262K context adds up to 2.6 GB (14.1 GB total).
VRAM = Weights + KV Cache + Overhead
Weights = 18.5B × 4.8 bits ÷ 8 = 11.1 GB
KV Cache + Overhead ≈ 0.4 GB (at 2K context + ~0.3 GB framework)
Fit ratings and hardware model lists check this model with room for a 16K-token context, which needs a little more memory.
KV Cache + Overhead ≈ 3 GB (at full 262K context)
VRAM usage by quantization
Can NVIDIA GeForce RTX 4090 run KAT Coder V2.5 Dev REAP 50 NVFP4A16?
Yes, at Q8_0 (18.9 GB) or lower. Higher quantizations like BF16 (37.5 GB) exceed the NVIDIA GeForce RTX 4090's 24 GB.
What's the best quantization for KAT Coder V2.5 Dev REAP 50 NVFP4A16?
For KAT Coder V2.5 Dev REAP 50 NVFP4A16, Q4_K_M (11.5 GB) offers the best balance of quality and VRAM usage. Q5_K_M (13.6 GB) provides better quality if you have the VRAM. The smallest option is Q2_K at 8.3 GB.
VRAM requirement by quantization
★ Recommended — best balance of quality and VRAM usage.
Can I run KAT Coder V2.5 Dev REAP 50 NVFP4A16 on a Mac?
Yes — Mac Mini M2 Pro (16 GB) and 145 other Macs can run KAT Coder V2.5 Dev REAP 50 NVFP4A16. Apple Silicon uses unified memory, so the model shares RAM with the system. At Q4_K_M you need at least 11.5 GB of usable unified memory (RAM minus macOS overhead).
Can I run KAT Coder V2.5 Dev REAP 50 NVFP4A16 locally?
Yes — KAT Coder V2.5 Dev REAP 50 NVFP4A16 can run locally on consumer hardware. Popular tools include Ollama, LM Studio, and llama.cpp.
How fast is KAT Coder V2.5 Dev REAP 50 NVFP4A16?
At Q4_K_M, KAT Coder V2.5 Dev REAP 50 NVFP4A16 can reach ~378 tok/s on AMD Instinct MI350X. On NVIDIA GeForce RTX 4090: ~183 tok/s. Speed depends mainly on GPU memory bandwidth. Real-world results typically within ±20%.
tok/s = 1000 ÷ (active GB ÷ (bandwidth GB/s × efficiency) × 1000 + layers × routing ms)
Mixture-of-Experts: just the active experts get read for each token, along with a fixed routing cost per layer.
Example: NVIDIA B200 → 2.1 GB active ÷ (8000 × 0.65) = 0.41 ms, plus 40 layers × 0.055 ms = 2.20 ms, so 1000 ÷ 2.61 ms = ~383 tok/s
Estimated speed at Q4_K_M (11.5 GB)
Real-world results typically within ±20%. Speed depends on batch size, quantization kernel, and software stack.
What's the download size of KAT Coder V2.5 Dev REAP 50 NVFP4A16?
At Q4_K_M, the download is about 11.13 GB. The full-precision BF16 version is 37.09 GB. The smallest option (Q2_K) is 7.88 GB.
Which GPUs can run KAT Coder V2.5 Dev REAP 50 NVFP4A16?
46 consumer GPUs can run KAT Coder V2.5 Dev REAP 50 NVFP4A16 at Q4_K_M. Top options include AMD Radeon RX 7900 XT, AMD Radeon RX 7900 XTX, NVIDIA GeForce RTX 3090, AMD Radeon RX 6700 XT. 8 GPUs have plenty of headroom for comfortable inference.
Which devices can run KAT Coder V2.5 Dev REAP 50 NVFP4A16?
169 devices with unified memory can run KAT Coder V2.5 Dev REAP 50 NVFP4A16 at Q4_K_M (11.5 GB), including AMD Ryzen 7 8845HS (Hawk Point) Laptop, AMD Ryzen AI 7 350 (Krackan Point) Laptop, AMD Ryzen AI 9 HX 370 (Strix Point) Laptop, ASUS Ascent GX10. Apple Silicon Macs use unified memory shared between CPU and GPU, making them well-suited for local LLM inference.