Can 8GB VRAM class (RTX 4060 8GB, 32GB RAM) run LFM3 1.7B?
Estimated speed and memory fit for LFM3 1.7B (1.73B dense) on 8GB VRAM class (RTX 4060 8GB, 32GB RAM). We have no measured run for this pair yet, so the speed is an estimate; submit yours.
Short answer
Yes. LFM3 1.7B Q8_0 fits fully on 8GB VRAM class (RTX 4060 8GB, 32GB RAM) at 16k context, using 2.1 of 7.0 GB usable, and generates about ~114 tokens/s (estimated, not measured). For Q8_0, our bandwidth-only estimate was ~81 tok/s against a measured median of ~114 (29% too pessimistic).
Measured generation speed by quantization
No measured rows for LFM3 1.7B on 8GB VRAM class (RTX 4060 8GB, 32GB RAM) yet. The ~114 tok/s above is a bandwidth estimate (272 GB/s × 0.95 efficiency). Have this machine? Submit a llama-bench result and this page will switch to measured data.
Generation = tg128 (128 generated tokens, short context). Public rows are medians of community llama.cpp results (sources); our own runs use llama-bench on a rented GPU.
Alternatives
Faster for LFM3 1.7B
- RTX 5070 12GB: Q8_0, 11 GB usable, ~228 tok/s scaled from measured
- RTX 4070 12GB: Q8_0, 11 GB usable, ~191 tok/s scaled from measured
- RTX 5080 16GB: Q8_0, 15 GB usable, ~320 tok/s scaled from measured
Other models on 8GB VRAM class (RTX 4060 8GB, 32GB RAM), similar size
- Nemotron 3 Nano 4B (hardware page): Q8_0, fits fully, ~54 tok/s (scaled from measured)
- Qwen3.5 4B: Q8_0, fits fully, ~51 tok/s (measured)
- DeepSeek R1 Distill Qwen 7B (hardware page): Q5_K_M, fits fully, ~48 tok/s (scaled from measured)
- Gemma 4 E4B (hardware page): Q5_K_M, fits fully, ~47 tok/s (scaled from measured)
Memory and context limits
Q8_0 weights are 1.9 GB; the 16k f16 KV cache adds 0.20 GB. Largest context that fits in 7.0 GB usable:
| KV cache type | Max context |
|---|---|
| f16 | 256k |
| q8_0 | 256k |
| q4_0 | 256k |
Native limit 256k. Full KV cache table for LFM3 1.7B.
Copy-paste commands
The advisor's pick for this hardware: Q8_0, 16k context.
llama-server -hf LiquidAI/LFM3-1.7B-GGUF:Q8_0 -c 16384 -fa on --jinja -ngl 99
OLLAMA_CONTEXT_LENGTH=16384 ollama serve ollama run hf.co/LiquidAI/LFM3-1.7B-GGUF:Q8_0
FAQ
How fast is LFM3 1.7B on 8GB VRAM class (RTX 4060 8GB, 32GB RAM)?
About 114.2 tokens/s for generation (Q8_0, estimated from 272 GB/s memory bandwidth). Speed drops as the context fills; see the depth table on this hardware's other pages.
How much memory does LFM3 1.7B need on 8GB VRAM class (RTX 4060 8GB, 32GB RAM)?
The advisor's pick, Q8_0, is 1.9 GB of weights plus 0.20 GB of KV cache at 16k context (2.1 GB), against 7.0 GB usable.
Is the LFM3 1.7B speed on 8GB VRAM class (RTX 4060 8GB, 32GB RAM) measured or estimated?
Estimated. No measured llama.cpp run for this pair is on file yet, so the figure comes from memory bandwidth and our calibration against measured hardware.
What can I use instead of 8GB VRAM class (RTX 4060 8GB, 32GB RAM) for LFM3 1.7B?
Nothing with less memory in our list fits it fully. Faster: RTX 5070 12GB (~228 tok/s), RTX 4070 12GB (~191 tok/s), RTX 5080 16GB (~320 tok/s).
What is the maximum context for LFM3 1.7B on 8GB VRAM class (RTX 4060 8GB, 32GB RAM)?
With an f16 KV cache about 256k tokens, 256k with q8_0 and 256k with q4_0, bounded by the model's own limit.
Related
All models on 8GB VRAM class (RTX 4060 8GB, 32GB RAM) · LFM3 1.7B on all hardware · VRAM tier · all measured pages · leaderboard
Compare: RTX 3060 12GB (32GB RAM) vs 8GB VRAM class (RTX 4060 8GB, 32GB RAM)