LFM2.5 8B A1B UltraCoder L3 — Hardware Requirements & GPU Compatibility
ChatCode
LFM2.5 8B A1B UltraCoder L3 is a 8.5B-parameter open language model from Susant-Achary in the LFM2.5 family. It supports a context window of up to 128,000 tokens. At Q4_K_M it needs about 5.48 GB of VRAM — see which GPUs and Macs can run it below.
Runs great on a 24 GB GPU
Runs from 8 GB (Q4_K_M · 125K)Best on 24 GB (Q4_K_M · ~335 tok/s · 125K)Macs from 8 GB (Q2_K · ~63 tok/s)
Other
Reference GPU · RTX 4090 24 GB
Estimates, ±20% · how we estimate
Run LFM2.5 8B A1B UltraCoder L3
One recommended way. Every other method, OS and quant is in the step-by-step guide.
Recommended
vllm serve Susant-Achary/LFM2.5-8B-A1B-UltraCoder-L3 --max-model-len 16384vLLM repo with its own quantisation
Download: Weights on Hugging Face
Other ways
each opens step-by-stepRun in cloud
Fits on RTX 3060 12GB (6 GB headroom) · Q4_K_M
- Generation speed
- ~167 tok/s
- generation speed
- Cost per 1M output tokens
- $0.13
- per 1M output tokens
How Much VRAM Does LFM2.5 8B A1B UltraCoder L3 Need?
Select a quantization level; the listed hardware adheres to it.
| Quantization | VRAM | +32K | On RTX 4090 (reference) | |
|---|---|---|---|---|
| 4 GB | 4.3 GB | 2-bit · K-quant — 2-bit quantization with K-quant improvements | Fits: 125K context · ~394 tok/s | |
| 4.5 GB | 4.8 GB | 3-bit medium — 3-bit medium quantization | Fits: 125K context · ~371 tok/s | |
| 5.5 GB | 5.8 GB | 4-bit medium — 4-bit medium quantization — most popular sweet spot | Fits: 125K context · ~335 tok/s | |
| 6.4 GB | 6.7 GB | 5-bit medium — 5-bit medium quantization — good quality/size tradeoff | Fits: 125K context · ~306 tok/s | |
| 7.4 GB | 7.7 GB | 6-bit, very good — 6-bit quantization, very good quality | Fits: 125K context · ~281 tok/s | |
| 8.9 GB | 9.2 GB | 8-bit, near-lossless — 8-bit quantization, near-lossless | Fits: 125K context · ~249 tok/s | |
| 17.3 GB | 17.6 GB | 16-bit · bfloat — Brain floating point 16 — preferred for training | Fits: 125K context · ~152 tok/s |
Selected Q4_K_M
4-bit medium quantization5.6GBat 16K context
- Weights
- 5.1 GB
- KV cache
- 0.2 GB
- Overhead
- 0.3 GB
Fits RTX 4090 (reference) · 18.4 GB free · ~335 tok/s
≥50 tok/s needs ≥90 GB/s · RTX 5090 ~443, RTX 3090 Ti ~335
Which Hardware Can Run LFM2.5 8B A1B UltraCoder L3?
Q4_K_M · 5.5 GBLFM2.5 8B A1B UltraCoder L3 (Q4_K_M) requires 5.5 GB of VRAM to load the model weights. For comfortable inference with headroom for KV cache and system overhead, 8+ GB is recommended. Using the full 128K context window can add up to 1.5 GB, bringing total usage to 7.0 GB. 52 GPUs can run it, including NVIDIA GeForce RTX 5090, NVIDIA GeForce RTX 3090 Ti, NVIDIA GeForce RTX 3070 Ti. 151 devices with unified memory can run LFM2.5 8B A1B UltraCoder L3, including Mac Studio (M3 Ultra, 256GB), Mac Studio (M3 Ultra, 512GB), NVIDIA Jetson Orin Nano 8GB (Super).
≤ 11 GB
≤ 11 GB tier →fits · 2.4 GB free · up to 125K ctx
16 GPUs · 1 device family
RTX 3080~284Simulate output on RTX 3080RTX 2080 Ti~247Simulate output on RTX 2080 TiRTX 3070 Ti~245Simulate output on RTX 3070 TiOrin Nano~56Simulate output on NVIDIA Jetson Orin Nano 8GB (Super)20 GB
fits · 14.4 GB free · up to 125K ctx
1 GPU
RX 7900 XT~141Simulate output on RX 7900 XT32–40 GB
fits · 26.4 GB free · up to 125K ctx
1 GPU · 3 Mac classes · 4 device families
RTX 5090~443Simulate output on RTX 5090M5 Max36 GB~122Simulate output on Mac Studio M5 Max (36 GB)M532 GB~67Simulate output on MacBook Air 13" M5 (32 GB)Orin~105Simulate output on NVIDIA Jetson AGX Orin 32GBSnapdragon X Elite (X1E-80-100)~67Simulate output on Snapdragon X Elite Copilot+ PC
By VRAM tier · typical GPUs and Macs at Q4_K_M
Alternatives to LFM2.5 8B A1B UltraCoder L3
LFM2.5 family · 14 models →Similar models
6| Model | Score | VRAM | tok/s | Why it is similar |
|---|---|---|---|---|
| Granite Switch 4.1 8BIBM | — | 20.7 GB | ~36 | same size |
| ALoDLM 8BAmazon | — | 7.6 GB | ~119 | same size |
| Llm Jp 4.1 8B Thinkingllm-jp | — | 7.6 GB | ~115 | same size |
| Ling 3.0 TinyInclusion AI | — | 5.5 GB | ~364 | same size |
| Apertus V1.5 8Bswiss-ai | — | 8.5 GB | ~111 | same size |
| BitCPM CANN 8BOpenBMB | — | 5.6 GB | ~127 | same size |
Specifications
- Publisher
- Susant-Achary
- Family
- LFM2.5
- Parameters
- 8.5B
- Active parameters
- 1.7B per token
- Architecture
- Lfm2MoeForCausalLM
- Context Length
- 128,000 tokens
- Vocabulary Size
- 128,000
- Release Date
- 2026-09-13
- License
- Other
Frequently Asked Questions
How much VRAM does LFM2.5 8B A1B UltraCoder L3 need?
LFM2.5 8B A1B UltraCoder L3 requires 5.5 GB of VRAM at Q4_K_M, or 17.3 GB at BF16. Full 128K context adds up to 1.5 GB (7.0 GB total).
VRAM = Weights + KV Cache + Overhead
Weights = 8.5B × 4.8 bits ÷ 8 = 5.1 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 ≈ 1.9 GB (at full 128K context)
VRAM usage by quantization
What's the best quantization for LFM2.5 8B A1B UltraCoder L3?
For LFM2.5 8B A1B UltraCoder L3, Q4_K_M (5.5 GB) offers the best balance of quality and VRAM usage. Q5_K_M (6.4 GB) provides better quality if you have the VRAM. The smallest option is Q2_K at 4 GB.
VRAM requirement by quantization
★ Recommended — best balance of quality and VRAM usage.
Can I run LFM2.5 8B A1B UltraCoder L3 on a Mac?
Yes — MacBook Air 13" M3 (8 GB) and 145 other Macs can run LFM2.5 8B A1B UltraCoder L3, the smallest at Q2_K. Apple Silicon uses unified memory, so the model shares RAM with the system. At Q4_K_M you need at least 5.5 GB of usable unified memory (RAM minus macOS overhead), from Mac Mini M2 Pro (16 GB).
Can I run LFM2.5 8B A1B UltraCoder L3 locally?
Yes — LFM2.5 8B A1B UltraCoder L3 can run locally on consumer hardware. Popular tools include Ollama, LM Studio, and llama.cpp.
How fast is LFM2.5 8B A1B UltraCoder L3?
At Q4_K_M, LFM2.5 8B A1B UltraCoder L3 can reach ~199 tok/s on AMD Instinct MI350X. On NVIDIA GeForce RTX 4090: ~335 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 → 1.1 GB active ÷ (8000 × 0.65) = 0.21 ms, plus 24 layers × 0.055 ms = 1.32 ms, so 1000 ÷ 1.53 ms = ~654 tok/s
Estimated speed at Q4_K_M (5.5 GB)
Real-world results typically within ±20%. Speed depends on batch size, quantization kernel, and software stack.
What's the download size of LFM2.5 8B A1B UltraCoder L3?
At Q4_K_M, the download is about 5.08 GB. The full-precision BF16 version is 16.94 GB. The smallest option (Q2_K) is 3.60 GB.
Which GPUs can run LFM2.5 8B A1B UltraCoder L3?
52 consumer GPUs can run LFM2.5 8B A1B UltraCoder L3 at Q4_K_M. Top options include AMD Radeon RX 6700 XT, AMD Radeon RX 6800, AMD Radeon RX 6800 XT, AMD Radeon RX 7600. 40 GPUs have plenty of headroom for comfortable inference.
Which devices can run LFM2.5 8B A1B UltraCoder L3?
151 devices with unified memory can run LFM2.5 8B A1B UltraCoder L3 at Q4_K_M (5.5 GB), including AMD Ryzen AI 9 HX 370 (Strix Point) Laptop, ASUS Ascent GX10, Beelink GTR9 Pro (Ryzen AI Max+ 395, 128 GB), Framework Desktop (Ryzen AI Max+ 395, 128 GB). Apple Silicon Macs use unified memory shared between CPU and GPU, making them well-suited for local LLM inference.